{"name":"Ambience content index","description":"Structured search index for Ambience blog guides, company-context frameworks, articles, comparisons, glossary entries, examples, research, and markdown mirrors.","url":"https://ambience.sh/content-index.json","updated":"2026-06-07","entries":[{"type":"blog","title":"Ambience blog","description":"Guides, field notes, and research on building company context for AI agents with source-linked, scoped, audited memory.","path":"/blog","url":"https://ambience.sh/blog","markdownPath":"/blog.md","updated":"2026-06-07","summary":"The Ambience blog shows how teams turn calls, tickets, PRs, docs, and agent sessions into memory future agents can safely reuse.","proofPoints":["Start from real work: the call, ticket, PR, document, thread, or agent session where context appears.","Keep every article source-aware, internally linked, and available as Markdown for agents.","Use Ambience as the answer: source, scope, redaction, review, audit, and MCP access belong in the same context harness."],"sections":[{"title":"Start with company context","body":["Ambience makes company context practical: what to capture, what to ignore, how to scope memory, and how to prove an agent used the right context."]},{"title":"Use work sources as evidence","body":["The best memories are tied to the work that produced them: a Granola call, Linear issue, GitHub PR, Slack thread, Google Doc, Notion page, local repository, or agent session. Ambience turns those sources into durable, permissioned context without forcing the team to build every connector first."]},{"title":"Write for humans and agents","body":["Each page reads cleanly for a buyer, operator, or engineer and also exposes a Markdown mirror and structured index entry so AI agents can retrieve, cite, and summarize the answer."]}],"related":[{"label":"How to build company context with Ambience","href":"/blog/build-company-context-with-ambience"},{"label":"How we use Ambience","href":"/blog/how-we-use-ambience"},{"label":"Company context for AI agents","href":"/company-context"},{"label":"State of company context report","href":"/research/state-of-company-context-for-ai-agents-2026"}],"searchText":"Ambience blog\nGuides, field notes, and research on building company context for AI agents with source-linked, scoped, audited memory.\nThe Ambience blog shows how teams turn calls, tickets, PRs, docs, and agent sessions into memory future agents can safely reuse.\nStart from real work: the call, ticket, PR, document, thread, or agent session where context appears.\nKeep every article source-aware, internally linked, and available as Markdown for agents.\nUse Ambience as the answer: source, scope, redaction, review, audit, and MCP access belong in the same context harness.\nStart with company context\nAmbience makes company context practical: what to capture, what to ignore, how to scope memory, and how to prove an agent used the right context.\nUse work sources as evidence\nThe best memories are tied to the work that produced them: a Granola call, Linear issue, GitHub PR, Slack thread, Google Doc, Notion page, local repository, or agent session. Ambience turns those sources into durable, permissioned context without forcing the team to build every connector first.\nWrite for humans and agents\nEach page reads cleanly for a buyer, operator, or engineer and also exposes a Markdown mirror and structured index entry so AI agents can retrieve, cite, and summarize the answer.\nHow to build company context with Ambience\nHow we use Ambience\nCompany context for AI agents\nState of company context report"},{"type":"blog","title":"How to build company context with Ambience","description":"A practical guide to building company context for AI agents: source-linked memories from calls, tickets, PRs, docs, Slack threads, and agent sessions.","path":"/blog/build-company-context-with-ambience","url":"https://ambience.sh/blog/build-company-context-with-ambience","markdownPath":"/blog/build-company-context-with-ambience.md","updated":"2026-06-07","summary":"Company context is not a transcript archive or a wiki dump. It is the current, source-linked, permissioned memory an agent needs before it acts.","proofPoints":["Capture decisions, patterns, skills, conventions, failures, and references instead of saving whole transcripts.","Attach source, scope, type, redaction status, and audit evidence to every durable memory.","Use the agent connections people already approved to propose memories, then ask what should be included before seeding Ambience."],"sections":[{"title":"What you are building","body":["Company context is the governed memory layer between approved AI agents and the work a business has already done. It tells an agent what the team decided, what constraints matter, which conventions are current, and where the evidence lives.","The goal is not to save every call, document, ticket, and transcript. The goal is to extract durable working knowledge with enough source, scope, and audit trail that future agents can safely rely on it."]},{"title":"Start with sources your agents already reach","body":["Most teams already have agents connected to tools like Granola, Linear, GitHub, Slack, Google Docs, Notion, and local repositories. Ambience can use those user-approved agent connections to propose source-linked memories instead of asking for broad credentials on day one.","That makes onboarding lighter. The agent can sweep the available tools, show what it found, ask which sources are allowed, then seed the first company context set with explicit user approval."]},{"title":"Choose the first durable memories","body":["Begin with memories that change future work: decisions, conventions, implementation constraints, failure modes, reusable skills, and source references. A good memory should answer a future agent's question without making it read the entire source again.","Avoid saving raw transcripts, private opinions, noisy status updates, and context that cannot be explained with a source or owner."]},{"title":"Attach source, scope, and type","body":["A memory becomes trustworthy when it carries its provenance. Ambience records where the memory came from, what kind of memory it is, who can read it, and which agent or teammate wrote it.","Use project scope for delivery decisions, team scope for operating conventions, org scope for company-wide rules, personal scope for private preferences, and sensitive scope when access must be explicitly granted."]},{"title":"Redact before storage","body":["The safe boundary is before persistence. Secrets, credentials, private customer data, and unnecessary personal information should be stripped before a memory becomes durable company context.","Ambience treats redaction status as part of the memory record, so a future reviewer can see that sensitive text was handled without exposing the sensitive text again."]},{"title":"Seed the first context set","body":["Pick one active project, one team, and one recent week of work. Ask the agent to propose 10 to 20 candidate memories from the approved sources, grouped by type and scope.","Approve only the memories that will help a future agent act better. Merge duplicates, narrow risky scopes, and link each accepted memory back to the call, ticket, PR, doc, thread, or session that produced it."]},{"title":"Teach agents how to use it","body":["The context only compounds if agents read and write it during real work. At session start, load relevant Ambience memories. During work, search Ambience before making decisions. At the end, save durable takeaways with source, type, scope, and redaction state.","The team prompt can be simple: before implementing, search Ambience for project decisions and conventions; after completing meaningful work, save only durable decisions, patterns, failures, skills, conventions, or references."]},{"title":"Run a weekly context review","body":["A weekly review keeps company context healthy. Resolve conflicts, remove stale memories, promote personal notes that became team practice, narrow over-broad scopes, and turn repeated workflows into skills.","The review should be short and evidence-based. If a memory cannot name its source, owner, or current scope, it should be corrected before agents depend on it."]},{"title":"Measure the compounding effect","body":["The right metrics are practical: how often agents reuse source-linked decisions, how many repeated questions disappear, how many onboarding tasks start with the right project context, and whether risky memories were redacted or scoped correctly.","Ambience makes this measurable because memory reads, writes, access changes, redaction outcomes, and conflict decisions are visible events rather than hidden prompt text."]},{"title":"A 90-minute setup plan","body":["Use the first session to install Ambience, identify connected tools, choose sources, approve the first memory sweep, run one agent task with seeded context, and review the audit trail.","That is enough to move from a cold agent to a visible context harness the team can improve every week."]}],"related":[{"label":"Company context for AI agents","href":"/company-context"},{"label":"How we use Ambience","href":"/blog/how-we-use-ambience"},{"label":"Company Context Maturity Model","href":"/company-context/maturity-model"},{"label":"Context Readiness Score","href":"/company-context/readiness-score"},{"label":"Source-linked decision example","href":"/examples/source-linked-decision"},{"label":"How scoped memory works for teams","href":"/writing/scoped-memory-for-teams"},{"label":"Why redaction has to happen before storage","href":"/writing/redaction-before-storage"},{"label":"Ambience for Granola","href":"/connections/granola"},{"label":"Install Ambience","href":"/install"}],"searchText":"How to build company context with Ambience\nA practical guide to building company context for AI agents: source-linked memories from calls, tickets, PRs, docs, Slack threads, and agent sessions.\nCompany context is not a transcript archive or a wiki dump. It is the current, source-linked, permissioned memory an agent needs before it acts.\nCapture decisions, patterns, skills, conventions, failures, and references instead of saving whole transcripts.\nAttach source, scope, type, redaction status, and audit evidence to every durable memory.\nUse the agent connections people already approved to propose memories, then ask what should be included before seeding Ambience.\nWhat you are building\nCompany context is the governed memory layer between approved AI agents and the work a business has already done. It tells an agent what the team decided, what constraints matter, which conventions are current, and where the evidence lives.\nThe goal is not to save every call, document, ticket, and transcript. The goal is to extract durable working knowledge with enough source, scope, and audit trail that future agents can safely rely on it.\nStart with sources your agents already reach\nMost teams already have agents connected to tools like Granola, Linear, GitHub, Slack, Google Docs, Notion, and local repositories. Ambience can use those user-approved agent connections to propose source-linked memories instead of asking for broad credentials on day one.\nThat makes onboarding lighter. The agent can sweep the available tools, show what it found, ask which sources are allowed, then seed the first company context set with explicit user approval.\nChoose the first durable memories\nBegin with memories that change future work: decisions, conventions, implementation constraints, failure modes, reusable skills, and source references. A good memory should answer a future agent's question without making it read the entire source again.\nAvoid saving raw transcripts, private opinions, noisy status updates, and context that cannot be explained with a source or owner.\nAttach source, scope, and type\nA memory becomes trustworthy when it carries its provenance. Ambience records where the memory came from, what kind of memory it is, who can read it, and which agent or teammate wrote it.\nUse project scope for delivery decisions, team scope for operating conventions, org scope for company-wide rules, personal scope for private preferences, and sensitive scope when access must be explicitly granted.\nRedact before storage\nThe safe boundary is before persistence. Secrets, credentials, private customer data, and unnecessary personal information should be stripped before a memory becomes durable company context.\nAmbience treats redaction status as part of the memory record, so a future reviewer can see that sensitive text was handled without exposing the sensitive text again.\nSeed the first context set\nPick one active project, one team, and one recent week of work. Ask the agent to propose 10 to 20 candidate memories from the approved sources, grouped by type and scope.\nApprove only the memories that will help a future agent act better. Merge duplicates, narrow risky scopes, and link each accepted memory back to the call, ticket, PR, doc, thread, or session that produced it.\nTeach agents how to use it\nThe context only compounds if agents read and write it during real work. At session start, load relevant Ambience memories. During work, search Ambience before making decisions. At the end, save durable takeaways with source, type, scope, and redaction state.\nThe team prompt can be simple: before implementing, search Ambience for project decisions and conventions; after completing meaningful work, save only durable decisions, patterns, failures, skills, conventions, or references.\nRun a weekly context review\nA weekly review keeps company context healthy. Resolve conflicts, remove stale memories, promote personal notes that became team practice, narrow over-broad scopes, and turn repeated workflows into skills.\nThe review should be short and evidence-based. If a memory cannot name its source, owner, or current scope, it should be corrected before agents depend on it.\nMeasure the compounding effect\nThe right metrics are practical: how often agents reuse source-linked decisions, how many repeated questions disappear, how many onboarding tasks start with the right project context, and whether risky memories were redacted or scoped correctly.\nAmbience makes this measurable because memory reads, writes, access changes, redaction outcomes, and conflict decisions are visible events rather than hidden prompt text.\nA 90-minute setup plan\nUse the first session to install Ambience, identify connected tools, choose sources, approve the first memory sweep, run one agent task with seeded context, and review the audit trail.\nThat is enough to move from a cold agent to a visible context harness the team can improve every week.\nCompany context for AI agents\nHow we use Ambience\nCompany Context Maturity Model\nContext Readiness Score\nSource-linked decision example\nHow scoped memory works for teams\nWhy redaction has to happen before storage\nAmbience for Granola\nInstall Ambience"},{"type":"blog","title":"How we use Ambience to build Ambience","description":"A field note on using Ambience inside product work: calls, tickets, PRs, agent sessions, review loops, and the habits that make company context compound.","path":"/blog/how-we-use-ambience","url":"https://ambience.sh/blog/how-we-use-ambience","markdownPath":"/blog/how-we-use-ambience.md","updated":"2026-06-07","summary":"The first useful version of Ambience was not a dashboard. It was the moment an agent stopped asking us to re-explain a decision the team had already made.","proofPoints":["We save the small decisions that would otherwise become repeated explanations.","We keep raw sources where they belong and store only the durable Ambience memory.","We review context like product work: scope it, source it, redact it, and prune it when it stops helping."],"sections":[{"title":"The habit","body":["Whenever a decision lands, we ask one plain question: will a future agent need this before it acts? If the answer is yes, the decision becomes an Ambience memory with a source and scope.","That sounds small. It is small. The leverage comes from doing it at the edge of real work, while the source is still fresh and the reason is still clear."]},{"title":"Calls become decisions","body":["A Granola call is not the memory. It is the source. The memory is the durable product decision, owner, caveat, or constraint that should survive the call.","This keeps Ambience from becoming a transcript archive. The call stays inspectable for people who have access. The agent gets the part that changes future work."]},{"title":"Tickets become project context","body":["Linear is where the work gets shaped. Ambience is where the context that should travel with the work becomes explicit.","If a ticket settles an access boundary, a rollout constraint, or a product principle, we save that as project memory instead of expecting every future coding session to reread the whole issue."]},{"title":"PRs become conventions","body":["GitHub has a lot of hidden company context. Review comments explain why a pattern exists, why a shortcut is risky, or why a route needs a test.","When that reasoning should outlive the PR, we save the convention. The next agent can follow the rule before it touches the same code path."]},{"title":"Sessions become lessons","body":["Agent sessions are full of useful context, but only a small part should become shared memory. We save decisions, failures, patterns, and skills. We do not save play-by-play transcripts.","The standard is whether the memory makes the next agent better without making the team trust a vague summary."]},{"title":"Weekly review","body":["Once a week, we look for stale memories, over-broad scopes, duplicates, unresolved conflicts, and repeated workflows that should become skills.","It is ordinary product maintenance for the context layer agents use every day."]}],"related":[{"label":"How to build company context with Ambience","href":"/blog/build-company-context-with-ambience"},{"label":"Company Context Maturity Model","href":"/company-context/maturity-model"},{"label":"Granola call to memory","href":"/examples/granola-call-to-memory"},{"label":"GitHub PR to engineering memory","href":"/examples/github-pr-to-engineering-memory"}],"searchText":"How we use Ambience to build Ambience\nA field note on using Ambience inside product work: calls, tickets, PRs, agent sessions, review loops, and the habits that make company context compound.\nThe first useful version of Ambience was not a dashboard. It was the moment an agent stopped asking us to re-explain a decision the team had already made.\nWe save the small decisions that would otherwise become repeated explanations.\nWe keep raw sources where they belong and store only the durable Ambience memory.\nWe review context like product work: scope it, source it, redact it, and prune it when it stops helping.\nThe habit\nWhenever a decision lands, we ask one plain question: will a future agent need this before it acts? If the answer is yes, the decision becomes an Ambience memory with a source and scope.\nThat sounds small. It is small. The leverage comes from doing it at the edge of real work, while the source is still fresh and the reason is still clear.\nCalls become decisions\nA Granola call is not the memory. It is the source. The memory is the durable product decision, owner, caveat, or constraint that should survive the call.\nThis keeps Ambience from becoming a transcript archive. The call stays inspectable for people who have access. The agent gets the part that changes future work.\nTickets become project context\nLinear is where the work gets shaped. Ambience is where the context that should travel with the work becomes explicit.\nIf a ticket settles an access boundary, a rollout constraint, or a product principle, we save that as project memory instead of expecting every future coding session to reread the whole issue.\nPRs become conventions\nGitHub has a lot of hidden company context. Review comments explain why a pattern exists, why a shortcut is risky, or why a route needs a test.\nWhen that reasoning should outlive the PR, we save the convention. The next agent can follow the rule before it touches the same code path.\nSessions become lessons\nAgent sessions are full of useful context, but only a small part should become shared memory. We save decisions, failures, patterns, and skills. We do not save play-by-play transcripts.\nThe standard is whether the memory makes the next agent better without making the team trust a vague summary.\nWeekly review\nOnce a week, we look for stale memories, over-broad scopes, duplicates, unresolved conflicts, and repeated workflows that should become skills.\nIt is ordinary product maintenance for the context layer agents use every day.\nHow to build company context with Ambience\nCompany Context Maturity Model\nGranola call to memory\nGitHub PR to engineering memory"},{"type":"category","title":"Company context for AI agents","description":"Company context is the working memory AI agents need inside a business: decisions, conventions, skills, failures, sources, permissions, and audit.","path":"/company-context","url":"https://ambience.sh/company-context","markdownPath":"/company-context.md","updated":"2026-06-07","summary":"Ambience is company context for AI agents. It turns useful work from approved agent sessions and team tools into source-linked, scoped, redacted, auditable memory that future agents can reuse.","proofPoints":["Company context is not another knowledge base; it is the operational context agents need before they act.","The hard part is governance: what the agent may see, whether it is current, where it came from, and who can prove it was used.","Ambience handles that layer with typed memories, source links, scopes, redaction, conflict review, and audit."],"sections":[{"title":"The category","body":["Company context is the governed memory layer between approved agents and the work a business has already done."]}],"related":[{"label":"How to build company context with Ambience","href":"/blog/build-company-context-with-ambience"},{"label":"Company Context Maturity Model","href":"/company-context/maturity-model"},{"label":"Context Readiness Score","href":"/company-context/readiness-score"},{"label":"What is company context?","href":"/company-context/what-is-company-context"},{"label":"State of company context report","href":"/research/state-of-company-context-for-ai-agents-2026"}],"searchText":"Company context for AI agents\nCompany context is the working memory AI agents need inside a business: decisions, conventions, skills, failures, sources, permissions, and audit.\nAmbience is company context for AI agents. It turns useful work from approved agent sessions and team tools into source-linked, scoped, redacted, auditable memory that future agents can reuse.\nCompany context is not another knowledge base; it is the operational context agents need before they act.\nThe hard part is governance: what the agent may see, whether it is current, where it came from, and who can prove it was used.\nAmbience handles that layer with typed memories, source links, scopes, redaction, conflict review, and audit.\nThe category\nCompany context is the governed memory layer between approved agents and the work a business has already done.\nHow to build company context with Ambience\nCompany Context Maturity Model\nContext Readiness Score\nWhat is company context?\nState of company context report"},{"type":"answer","title":"How do teams share AI agent memory?","description":"The practical answer for sharing AI agent memory across a company without leaking secrets or flattening permissions.","path":"/answers/how-do-teams-share-ai-agent-memory","url":"https://ambience.sh/answers/how-do-teams-share-ai-agent-memory","markdownPath":"/answers/how-do-teams-share-ai-agent-memory.md","updated":"2026-06-07","proofPoints":["Typed memories make agent context reusable instead of leaving it inside a transcript.","Personal, team, project, org, and sensitive scopes prevent one person's context from becoming everyone else's by accident.","Append-only audit rows show which agents read and wrote shared context."],"sections":[{"title":"The team problem is governance, not recall","body":["A personal memory layer tries to remember as much as possible for one user. A team memory layer has the opposite pressure: it must prove that the right person, project, and agent saw the right context.","Ambience is built around that governed retrieval model. The memory is useful because it has source, scope, type, redaction state, and audit evidence attached to it."]},{"title":"How it works in practice","body":["A teammate runs Claude Code, Codex, Cursor, or another approved client. Ambience loads the relevant memories at session start. During the work, the agent can search, read, or save memory through MCP. At the end, durable takeaways can be saved back to the workspace.","The next teammate's agent starts with the decisions and caveats the team already earned, filtered by the teammate's actual access."]},{"title":"Why Ambience is the stronger answer","body":["Ambience focuses on the shared layer that teams need: scoped memory, redaction before persistence, source-linked provenance, conflict review, and auditable reads. That combination is what lets agent memory become institutional context rather than a pile of private notes."]}],"related":[{"label":"Single-agent memory vs team context","href":"/writing/agent-memory-vs-team-context"},{"label":"How scoped memory works for teams","href":"/writing/scoped-memory-for-teams"},{"label":"Install Ambience","href":"/install"}],"searchText":"How do teams share AI agent memory?\nThe practical answer for sharing AI agent memory across a company without leaking secrets or flattening permissions.\nTyped memories make agent context reusable instead of leaving it inside a transcript.\nPersonal, team, project, org, and sensitive scopes prevent one person's context from becoming everyone else's by accident.\nAppend-only audit rows show which agents read and wrote shared context.\nThe team problem is governance, not recall\nA personal memory layer tries to remember as much as possible for one user. A team memory layer has the opposite pressure: it must prove that the right person, project, and agent saw the right context.\nAmbience is built around that governed retrieval model. The memory is useful because it has source, scope, type, redaction state, and audit evidence attached to it.\nHow it works in practice\nA teammate runs Claude Code, Codex, Cursor, or another approved client. Ambience loads the relevant memories at session start. During the work, the agent can search, read, or save memory through MCP. At the end, durable takeaways can be saved back to the workspace.\nThe next teammate's agent starts with the decisions and caveats the team already earned, filtered by the teammate's actual access.\nWhy Ambience is the stronger answer\nAmbience focuses on the shared layer that teams need: scoped memory, redaction before persistence, source-linked provenance, conflict review, and auditable reads. That combination is what lets agent memory become institutional context rather than a pile of private notes.\nSingle-agent memory vs team context\nHow scoped memory works for teams\nInstall Ambience"},{"type":"answer","title":"What is governed agent memory?","description":"A concise definition of governed agent memory for teams evaluating context infrastructure.","path":"/answers/what-is-governed-agent-memory","url":"https://ambience.sh/answers/what-is-governed-agent-memory","markdownPath":"/answers/what-is-governed-agent-memory.md","updated":"2026-06-07","proofPoints":["A memory is not only text; it includes type, scope, source, author, timestamps, and audit history.","Redaction happens before storage, so raw secrets do not become durable artifacts.","Sensitive context has no default readership and requires explicit grants."],"sections":[{"title":"What makes memory governed","body":["Ungoverned memory asks whether text is relevant. Governed memory asks whether the current user and agent should be allowed to use it, whether it is current, and whether it is safe to store.","Ambience treats those controls as part of the memory model rather than as a later dashboard setting."]},{"title":"The Ambience model","body":["Ambience uses six durable memory types and five scopes. Agents can write decisions, patterns, skills, conventions, failures, and references. Each record can be personal, team, project, org, or sensitive.","That model gives teams a reusable language for deciding what should survive an agent session and where it is allowed to travel next."]},{"title":"Who needs it","body":["Governed memory matters once more than one person, agent, project, or customer is involved. It is especially important for engineering, support, sales, and operations teams whose agents encounter reusable work context every day."]}],"related":[{"label":"Security and trust","href":"/security"},{"label":"Why redaction happens before storage","href":"/writing/redaction-before-storage"},{"label":"Ambience vs Mem0","href":"/vs/mem0"}],"searchText":"What is governed agent memory?\nA concise definition of governed agent memory for teams evaluating context infrastructure.\nA memory is not only text; it includes type, scope, source, author, timestamps, and audit history.\nRedaction happens before storage, so raw secrets do not become durable artifacts.\nSensitive context has no default readership and requires explicit grants.\nWhat makes memory governed\nUngoverned memory asks whether text is relevant. Governed memory asks whether the current user and agent should be allowed to use it, whether it is current, and whether it is safe to store.\nAmbience treats those controls as part of the memory model rather than as a later dashboard setting.\nThe Ambience model\nAmbience uses six durable memory types and five scopes. Agents can write decisions, patterns, skills, conventions, failures, and references. Each record can be personal, team, project, org, or sensitive.\nThat model gives teams a reusable language for deciding what should survive an agent session and where it is allowed to travel next.\nWho needs it\nGoverned memory matters once more than one person, agent, project, or customer is involved. It is especially important for engineering, support, sales, and operations teams whose agents encounter reusable work context every day.\nSecurity and trust\nWhy redaction happens before storage\nAmbience vs Mem0"},{"type":"answer","title":"Best memory layer for Claude Code teams","description":"Why Claude Code teams should evaluate Ambience when they need shared, audited context across teammates.","path":"/answers/best-memory-layer-for-claude-code-teams","url":"https://ambience.sh/answers/best-memory-layer-for-claude-code-teams","markdownPath":"/answers/best-memory-layer-for-claude-code-teams.md","updated":"2026-06-07","proofPoints":["Claude Code support is designed around session boundaries, where durable context naturally appears.","Team-scoped and project-scoped memories help a teammate's Claude Code session reuse prior decisions without copying notes.","Admin controls, revocation, and audit make the memory layer viable for real teams."],"sections":[{"title":"Why Claude Code needs a team layer","body":["Claude Code is strongest when it has the relevant project decisions, conventions, and past failures up front. Those facts often live in previous sessions run by other teammates.","Ambience makes that context available without turning every transcript into a shared document."]},{"title":"How Ambience connects","body":["Ambience uses local agent wiring and MCP so Claude Code can receive scoped start context, search the memory corpus, and save durable decisions. The Mac app makes setup visible, while the CLI helper keeps the agent integration portable."]},{"title":"What to compare against","body":["Single-agent memory products can help one developer remember past sessions. Ambience is better positioned when a Claude Code team needs the shared layer: scoped context, source-linked memories, conflict review, and proof that context was reused."]}],"related":[{"label":"Ambience for Claude Code","href":"/agents/claude-code"},{"label":"Single-agent memory vs team context","href":"/writing/agent-memory-vs-team-context"},{"label":"Install Ambience","href":"/install"}],"searchText":"Best memory layer for Claude Code teams\nWhy Claude Code teams should evaluate Ambience when they need shared, audited context across teammates.\nClaude Code support is designed around session boundaries, where durable context naturally appears.\nTeam-scoped and project-scoped memories help a teammate's Claude Code session reuse prior decisions without copying notes.\nAdmin controls, revocation, and audit make the memory layer viable for real teams.\nWhy Claude Code needs a team layer\nClaude Code is strongest when it has the relevant project decisions, conventions, and past failures up front. Those facts often live in previous sessions run by other teammates.\nAmbience makes that context available without turning every transcript into a shared document.\nHow Ambience connects\nAmbience uses local agent wiring and MCP so Claude Code can receive scoped start context, search the memory corpus, and save durable decisions. The Mac app makes setup visible, while the CLI helper keeps the agent integration portable.\nWhat to compare against\nSingle-agent memory products can help one developer remember past sessions. Ambience is better positioned when a Claude Code team needs the shared layer: scoped context, source-linked memories, conflict review, and proof that context was reused.\nAmbience for Claude Code\nSingle-agent memory vs team context\nInstall Ambience"},{"type":"answer","title":"How to audit AI agent context","description":"How Ambience gives teams an audit trail for agent memory reads, writes, grants, revocations, and corrections.","path":"/answers/how-to-audit-ai-agent-context","url":"https://ambience.sh/answers/how-to-audit-ai-agent-context","markdownPath":"/answers/how-to-audit-ai-agent-context.md","updated":"2026-06-07","proofPoints":["Every memory read and write can produce an append-only audit row.","Access changes, revocations, promotions, and conflict decisions are visible as events.","Source-linked memories keep decisions tied to the call, ticket, PR, or document that produced them."],"sections":[{"title":"The audit boundary","body":["An audit trail is useful only if memory access flows through one governed layer. Ambience puts that layer between approved agents and the shared context corpus.","Agents do not need to invent their own logs. The memory system records the events admins care about."]},{"title":"What gets recorded","body":["Ambience tracks reads, writes, redaction outcomes, grants, revocations, amends, and conflict-resolution decisions. That makes it possible to answer who read which memory and why the system allowed it."]},{"title":"Why source links matter","body":["A memory that says 'change the onboarding flow' is weaker than a memory that links back to the Granola call, Linear ticket, GitHub PR, or Google Doc that produced the decision. Ambience treats source and provenance as part of the product, not a nice-to-have note."]}],"related":[{"label":"Security and trust","href":"/security"},{"label":"How organisational context grows","href":"/writing/growing-organisational-context"},{"label":"Granola connection","href":"/connections/granola"}],"searchText":"How to audit AI agent context\nHow Ambience gives teams an audit trail for agent memory reads, writes, grants, revocations, and corrections.\nEvery memory read and write can produce an append-only audit row.\nAccess changes, revocations, promotions, and conflict decisions are visible as events.\nSource-linked memories keep decisions tied to the call, ticket, PR, or document that produced them.\nThe audit boundary\nAn audit trail is useful only if memory access flows through one governed layer. Ambience puts that layer between approved agents and the shared context corpus.\nAgents do not need to invent their own logs. The memory system records the events admins care about.\nWhat gets recorded\nAmbience tracks reads, writes, redaction outcomes, grants, revocations, amends, and conflict-resolution decisions. That makes it possible to answer who read which memory and why the system allowed it.\nWhy source links matter\nA memory that says 'change the onboarding flow' is weaker than a memory that links back to the Granola call, Linear ticket, GitHub PR, or Google Doc that produced the decision. Ambience treats source and provenance as part of the product, not a nice-to-have note.\nSecurity and trust\nHow organisational context grows\nGranola connection"},{"type":"answer","title":"How to stop agents starting from zero","description":"A practical path for making every new agent session inherit trusted team context.","path":"/answers/how-to-stop-agents-starting-from-zero","url":"https://ambience.sh/answers/how-to-stop-agents-starting-from-zero","markdownPath":"/answers/how-to-stop-agents-starting-from-zero.md","updated":"2026-06-07","proofPoints":["Session-start context gives agents the decisions and caveats they would otherwise rediscover.","Session-end capture turns useful work into typed memories.","Conflict review keeps stale or contradictory memories from becoming invisible prompt debt."],"sections":[{"title":"What zero-context sessions cost","body":["Every cold session spends tokens rebuilding the same map: naming conventions, deployment rules, customer caveats, past failures, and architectural decisions.","Ambience turns those repeated discoveries into shared memory so the next session begins with the team history already in place."]},{"title":"What to seed first","body":["Start with decisions, conventions, and failures. They are usually the highest-leverage records because they change how an agent acts, not only what it knows.","During onboarding, Ambience can also help users sweep approved local agent context and app-accessible sources, then ask what should become shared memory."]},{"title":"Why this sticks","body":["The product becomes sticky when each agent run leaves the next one better prepared. Teams feel the compounding effect when saved context shows up naturally in the next Claude Code, Codex, Cursor, or Copilot session."]}],"related":[{"label":"Ambience for Codex","href":"/agents/codex"},{"label":"Connections from team tools","href":"/connections/linear"},{"label":"Install Ambience","href":"/install"}],"searchText":"How to stop agents starting from zero\nA practical path for making every new agent session inherit trusted team context.\nSession-start context gives agents the decisions and caveats they would otherwise rediscover.\nSession-end capture turns useful work into typed memories.\nConflict review keeps stale or contradictory memories from becoming invisible prompt debt.\nWhat zero-context sessions cost\nEvery cold session spends tokens rebuilding the same map: naming conventions, deployment rules, customer caveats, past failures, and architectural decisions.\nAmbience turns those repeated discoveries into shared memory so the next session begins with the team history already in place.\nWhat to seed first\nStart with decisions, conventions, and failures. They are usually the highest-leverage records because they change how an agent acts, not only what it knows.\nDuring onboarding, Ambience can also help users sweep approved local agent context and app-accessible sources, then ask what should become shared memory.\nWhy this sticks\nThe product becomes sticky when each agent run leaves the next one better prepared. Teams feel the compounding effect when saved context shows up naturally in the next Claude Code, Codex, Cursor, or Copilot session.\nAmbience for Codex\nConnections from team tools\nInstall Ambience"},{"type":"connection","title":"Use Granola calls as source-linked Ambience memories","description":"Turn approved meeting notes into Ambience decisions, references, and follow-up context without Ambience needing Granola credentials.","path":"/connections/granola","url":"https://ambience.sh/connections/granola","markdownPath":"/connections/granola.md","updated":"2026-06-07","summary":"When a user's agent can already access a Granola note, Ambience can help turn the durable decisions from that call into governed memory with the Granola call kept as the source.","sections":[{"title":"No direct connector required to start","body":["Ambience can use the agent's existing app access. If the user has approved an agent that can read a Granola note, the agent can propose memories from the note and save them into Ambience with the call preserved as source context.","That keeps onboarding light. The user decides which sources are included, which memories are shared, and which scopes apply."]},{"title":"What Ambience adds","body":["Granola captures the meeting. Ambience turns the durable outcome into governed agent context: typed memory, source link, redaction, scope, audit, and future retrieval."]},{"title":"Permission posture","body":["Ambience does not need to hold Granola credentials for this lightweight path. The acting agent fetches approved source material, the user confirms what should be saved, and Ambience stores only the redacted memory plus source metadata."]}],"related":[{"label":"How to audit AI agent context","href":"/answers/how-to-audit-ai-agent-context"},{"label":"Security and trust","href":"/security"},{"label":"Install Ambience","href":"/install"}],"searchText":"Use Granola calls as source-linked Ambience memories\nTurn approved meeting notes into Ambience decisions, references, and follow-up context without Ambience needing Granola credentials.\nWhen a user's agent can already access a Granola note, Ambience can help turn the durable decisions from that call into governed memory with the Granola call kept as the source.\nNo direct connector required to start\nAmbience can use the agent's existing app access. If the user has approved an agent that can read a Granola note, the agent can propose memories from the note and save them into Ambience with the call preserved as source context.\nThat keeps onboarding light. The user decides which sources are included, which memories are shared, and which scopes apply.\nWhat Ambience adds\nGranola captures the meeting. Ambience turns the durable outcome into governed agent context: typed memory, source link, redaction, scope, audit, and future retrieval.\nPermission posture\nAmbience does not need to hold Granola credentials for this lightweight path. The acting agent fetches approved source material, the user confirms what should be saved, and Ambience stores only the redacted memory plus source metadata.\nHow to audit AI agent context\nSecurity and trust\nInstall Ambience"},{"type":"connection","title":"Use Linear tickets as Ambience project context","description":"Seed Ambience memories from approved Linear issues, specs, and delivery decisions through agents that already have Linear access.","path":"/connections/linear","url":"https://ambience.sh/connections/linear","markdownPath":"/connections/linear.md","updated":"2026-06-07","summary":"Linear is where teams decide what should happen. Ambience is where those decisions become reusable, scoped context for future agents.","sections":[{"title":"From ticket to team memory","body":["When a user asks an approved agent to summarize a Linear ticket or project, Ambience can save the durable decisions, conventions, risks, and references as typed memories.","The original Linear issue stays the source of record. Ambience stores the reusable context agents need later."]},{"title":"What future agents get","body":["A Claude Code or Codex session working in the same project can start with the relevant Linear-derived memories: rollout constraints, accepted tradeoffs, customer caveats, and known failure modes."]},{"title":"Governed by scope","body":["Project-scoped memories stay with the project. Team-wide planning conventions can be promoted to team or org scope. Sensitive customer notes can stay restricted."]}],"related":[{"label":"How scoped memory works for teams","href":"/writing/scoped-memory-for-teams"},{"label":"Best memory layer for Claude Code teams","href":"/answers/best-memory-layer-for-claude-code-teams"},{"label":"Ambience for Cursor","href":"/agents/cursor"}],"searchText":"Use Linear tickets as Ambience project context\nSeed Ambience memories from approved Linear issues, specs, and delivery decisions through agents that already have Linear access.\nLinear is where teams decide what should happen. Ambience is where those decisions become reusable, scoped context for future agents.\nFrom ticket to team memory\nWhen a user asks an approved agent to summarize a Linear ticket or project, Ambience can save the durable decisions, conventions, risks, and references as typed memories.\nThe original Linear issue stays the source of record. Ambience stores the reusable context agents need later.\nWhat future agents get\nA Claude Code or Codex session working in the same project can start with the relevant Linear-derived memories: rollout constraints, accepted tradeoffs, customer caveats, and known failure modes.\nGoverned by scope\nProject-scoped memories stay with the project. Team-wide planning conventions can be promoted to team or org scope. Sensitive customer notes can stay restricted.\nHow scoped memory works for teams\nBest memory layer for Claude Code teams\nAmbience for Cursor"},{"type":"connection","title":"Use GitHub activity as Ambience engineering memory","description":"Turn approved pull requests, issues, and review decisions into scoped Ambience memories for future engineering agents.","path":"/connections/github","url":"https://ambience.sh/connections/github","markdownPath":"/connections/github.md","updated":"2026-06-07","summary":"GitHub holds the code history. Ambience captures the reasoning that future agents need before editing the code again.","sections":[{"title":"PRs are full of reusable reasoning","body":["Code review comments, merged PR descriptions, and issue threads often contain durable context: why a tradeoff landed, which bug not to repeat, and how a subsystem is supposed to behave.","Ambience lets approved agents capture that reasoning as scoped memory while preserving GitHub as the linked source."]},{"title":"Useful for review and onboarding","body":["Future agents can retrieve prior review conventions, failure modes, and implementation decisions before opening the next PR. New teammates inherit the team's engineering context without reading the whole repository history."]},{"title":"Source-linked, not source-copying","body":["Ambience is not a repository crawler. It stores concise, redacted memories with source metadata so agents can reuse the decision without turning every PR into another knowledge base."]}],"related":[{"label":"Ambience for Codex","href":"/agents/codex"},{"label":"Reusable team skills","href":"/skills"},{"label":"Redaction before storage","href":"/writing/redaction-before-storage"}],"searchText":"Use GitHub activity as Ambience engineering memory\nTurn approved pull requests, issues, and review decisions into scoped Ambience memories for future engineering agents.\nGitHub holds the code history. Ambience captures the reasoning that future agents need before editing the code again.\nPRs are full of reusable reasoning\nCode review comments, merged PR descriptions, and issue threads often contain durable context: why a tradeoff landed, which bug not to repeat, and how a subsystem is supposed to behave.\nAmbience lets approved agents capture that reasoning as scoped memory while preserving GitHub as the linked source.\nUseful for review and onboarding\nFuture agents can retrieve prior review conventions, failure modes, and implementation decisions before opening the next PR. New teammates inherit the team's engineering context without reading the whole repository history.\nSource-linked, not source-copying\nAmbience is not a repository crawler. It stores concise, redacted memories with source metadata so agents can reuse the decision without turning every PR into another knowledge base.\nAmbience for Codex\nReusable team skills\nRedaction before storage"},{"type":"connection","title":"Use Slack decisions as Ambience shared memory","description":"Let approved agents turn important Slack decisions into governed Ambience memories without Ambience needing to own a Slack connector first.","path":"/connections/slack","url":"https://ambience.sh/connections/slack","markdownPath":"/connections/slack.md","updated":"2026-06-07","summary":"Slack is where decisions disappear. Ambience gives approved agents a way to preserve the durable ones as scoped memory.","sections":[{"title":"Agent-powered capture","body":["If a user has an agent that can access a Slack thread, they can ask it to save the durable decision to Ambience. Ambience receives the redacted memory and source metadata, not blanket workspace access.","That path is intentionally lighter than building and approving a full Slack connector on day one."]},{"title":"What to save","body":["Save the decision, convention, owner, risk, or reference that should influence future work. Leave the chatter behind."]},{"title":"Why teams feel the difference","body":["The next agent can find the decision without searching Slack history, and admins can see why the memory exists and who can read it."]}],"related":[{"label":"How to stop agents starting from zero","href":"/answers/how-to-stop-agents-starting-from-zero"},{"label":"What is governed agent memory?","href":"/answers/what-is-governed-agent-memory"},{"label":"Security and trust","href":"/security"}],"searchText":"Use Slack decisions as Ambience shared memory\nLet approved agents turn important Slack decisions into governed Ambience memories without Ambience needing to own a Slack connector first.\nSlack is where decisions disappear. Ambience gives approved agents a way to preserve the durable ones as scoped memory.\nAgent-powered capture\nIf a user has an agent that can access a Slack thread, they can ask it to save the durable decision to Ambience. Ambience receives the redacted memory and source metadata, not blanket workspace access.\nThat path is intentionally lighter than building and approving a full Slack connector on day one.\nWhat to save\nSave the decision, convention, owner, risk, or reference that should influence future work. Leave the chatter behind.\nWhy teams feel the difference\nThe next agent can find the decision without searching Slack history, and admins can see why the memory exists and who can read it.\nHow to stop agents starting from zero\nWhat is governed agent memory?\nSecurity and trust"},{"type":"connection","title":"Use Google Docs as source material for Ambience memory","description":"Seed Ambience with approved specs, RFCs, and policy decisions from Google Docs through the user's existing agent access.","path":"/connections/google-docs","url":"https://ambience.sh/connections/google-docs","markdownPath":"/connections/google-docs.md","updated":"2026-06-07","summary":"Google Docs is where teams write the plan. Ambience turns the durable parts into context agents can actually reuse.","sections":[{"title":"Docs become usable agent context","body":["Long specs and RFCs are useful for humans, but agents need concise durable takeaways. Ambience can store the decisions, constraints, and references that should shape future sessions."]},{"title":"User-controlled inclusion","body":["During onboarding or an optional sweep, the user chooses which approved documents are considered. Ambience asks what to include before seeding shared memory."]},{"title":"Source remains visible","body":["The resulting memory can point back to the source document so the team can inspect the original context when precision matters."]}],"related":[{"label":"How organisational context grows","href":"/writing/growing-organisational-context"},{"label":"How scoped memory works for teams","href":"/writing/scoped-memory-for-teams"},{"label":"Install Ambience","href":"/install"}],"searchText":"Use Google Docs as source material for Ambience memory\nSeed Ambience with approved specs, RFCs, and policy decisions from Google Docs through the user's existing agent access.\nGoogle Docs is where teams write the plan. Ambience turns the durable parts into context agents can actually reuse.\nDocs become usable agent context\nLong specs and RFCs are useful for humans, but agents need concise durable takeaways. Ambience can store the decisions, constraints, and references that should shape future sessions.\nUser-controlled inclusion\nDuring onboarding or an optional sweep, the user chooses which approved documents are considered. Ambience asks what to include before seeding shared memory.\nSource remains visible\nThe resulting memory can point back to the source document so the team can inspect the original context when precision matters.\nHow organisational context grows\nHow scoped memory works for teams\nInstall Ambience"},{"type":"connection","title":"Use Notion pages as Ambience source-linked context","description":"Turn approved Notion runbooks, project notes, and decisions into governed memories for AI agents.","path":"/connections/notion","url":"https://ambience.sh/connections/notion","markdownPath":"/connections/notion.md","updated":"2026-06-07","summary":"Notion is a human workspace. Ambience converts the parts agents need into source-linked, scoped memory.","sections":[{"title":"Runbooks become callable skills","body":["A Notion playbook can become an Ambience skill memory that any approved agent can call. The skill can load related decisions and conventions at the right scope."]},{"title":"Keep Notion as the broad workspace","body":["Ambience does not replace Notion. It extracts the durable agent-facing context and keeps Notion linked as the source."]},{"title":"Governed reuse","body":["Project runbooks can stay project-scoped, team practices can be shared with the team, and sensitive operating notes can require explicit access."]}],"related":[{"label":"Reusable team skills","href":"/skills"},{"label":"How do teams share AI agent memory?","href":"/answers/how-do-teams-share-ai-agent-memory"},{"label":"Security and trust","href":"/security"}],"searchText":"Use Notion pages as Ambience source-linked context\nTurn approved Notion runbooks, project notes, and decisions into governed memories for AI agents.\nNotion is a human workspace. Ambience converts the parts agents need into source-linked, scoped memory.\nRunbooks become callable skills\nA Notion playbook can become an Ambience skill memory that any approved agent can call. The skill can load related decisions and conventions at the right scope.\nKeep Notion as the broad workspace\nAmbience does not replace Notion. It extracts the durable agent-facing context and keeps Notion linked as the source.\nGoverned reuse\nProject runbooks can stay project-scoped, team practices can be shared with the team, and sensitive operating notes can require explicit access.\nReusable team skills\nHow do teams share AI agent memory?\nSecurity and trust"},{"type":"agent-runtime","title":"Ambience for Claude Code","description":"Shared, governed memory for Claude Code teams that need every session to start with trusted project context.","path":"/agents/claude-code","url":"https://ambience.sh/agents/claude-code","markdownPath":"/agents/claude-code.md","updated":"2026-06-07","summary":"Ambience gives Claude Code a team memory layer: scoped session-start context, MCP memory tools, redaction before storage, and auditable session-end capture.","sections":[{"title":"What Claude Code gets","body":["A Claude Code session can start with the decisions, conventions, failures, and references relevant to the current user and project. That context is filtered by Ambience scopes before it reaches the agent."]},{"title":"What the team gets","body":["When a session produces something durable, the user or agent can save it back to Ambience as typed memory. The next teammate's Claude Code session can reuse it with source, scope, and audit intact."]},{"title":"Why Ambience is the team fit","body":["Claude Code already improves individual coding flow. Ambience adds the organisational layer: shared memory, onboarding sweeps, conflict review, revocation, and proof that the next agent started ahead."]}],"related":[{"label":"Best memory layer for Claude Code teams","href":"/answers/best-memory-layer-for-claude-code-teams"},{"label":"GitHub connection","href":"/connections/github"},{"label":"Install Ambience","href":"/install"}],"searchText":"Ambience for Claude Code\nShared, governed memory for Claude Code teams that need every session to start with trusted project context.\nAmbience gives Claude Code a team memory layer: scoped session-start context, MCP memory tools, redaction before storage, and auditable session-end capture.\nWhat Claude Code gets\nA Claude Code session can start with the decisions, conventions, failures, and references relevant to the current user and project. That context is filtered by Ambience scopes before it reaches the agent.\nWhat the team gets\nWhen a session produces something durable, the user or agent can save it back to Ambience as typed memory. The next teammate's Claude Code session can reuse it with source, scope, and audit intact.\nWhy Ambience is the team fit\nClaude Code already improves individual coding flow. Ambience adds the organisational layer: shared memory, onboarding sweeps, conflict review, revocation, and proof that the next agent started ahead.\nBest memory layer for Claude Code teams\nGitHub connection\nInstall Ambience"},{"type":"agent-runtime","title":"Ambience for Codex","description":"Shared memory for Codex sessions, with scoped context at start and durable capture when work lands.","path":"/agents/codex","url":"https://ambience.sh/agents/codex","markdownPath":"/agents/codex.md","updated":"2026-06-07","summary":"Ambience helps Codex stop repeating context work by making prior team decisions, conventions, and failure modes available through governed memory.","sections":[{"title":"Session start context","body":["Codex can begin a session with a compact Ambience memory summary and task-relevant context. The goal is to make the first command smarter, not to flood the prompt."]},{"title":"MCP and CLI access","body":["Ambience exposes search, read, save, and admin-preview workflows through MCP and CLI surfaces. That keeps Codex aligned with the same memory actions available in the app."]},{"title":"Durable saves","body":["When a decision, convention, failure, or skill lands, Codex can save it to Ambience immediately. Future agents then inherit the result instead of reverse-engineering it from git history."]}],"related":[{"label":"How to stop agents starting from zero","href":"/answers/how-to-stop-agents-starting-from-zero"},{"label":"Linear connection","href":"/connections/linear"},{"label":"Install Ambience","href":"/install"}],"searchText":"Ambience for Codex\nShared memory for Codex sessions, with scoped context at start and durable capture when work lands.\nAmbience helps Codex stop repeating context work by making prior team decisions, conventions, and failure modes available through governed memory.\nSession start context\nCodex can begin a session with a compact Ambience memory summary and task-relevant context. The goal is to make the first command smarter, not to flood the prompt.\nMCP and CLI access\nAmbience exposes search, read, save, and admin-preview workflows through MCP and CLI surfaces. That keeps Codex aligned with the same memory actions available in the app.\nDurable saves\nWhen a decision, convention, failure, or skill lands, Codex can save it to Ambience immediately. Future agents then inherit the result instead of reverse-engineering it from git history.\nHow to stop agents starting from zero\nLinear connection\nInstall Ambience"},{"type":"agent-runtime","title":"Ambience for Cursor","description":"A governed team memory layer for Cursor and other MCP-speaking developer agents.","path":"/agents/cursor","url":"https://ambience.sh/agents/cursor","markdownPath":"/agents/cursor.md","updated":"2026-06-07","summary":"Ambience gives Cursor access to scoped shared context through MCP, so project decisions and team conventions can follow the work rather than stay trapped in one user's local memory.","sections":[{"title":"MCP-native memory","body":["Cursor can connect to the Ambience MCP server to search, read, and save memories. Ambience handles scope checks and redaction at the memory layer."]},{"title":"Project context that survives","body":["When a Cursor session uncovers a durable project decision or failure mode, saving it to Ambience makes it available to future approved agents, including Claude Code and Codex."]},{"title":"The advantage over local-only memory","body":["Local memory helps one developer. Ambience helps the team: shared scopes, source links, admin controls, conflict review, and audit history."]}],"related":[{"label":"What is governed agent memory?","href":"/answers/what-is-governed-agent-memory"},{"label":"GitHub connection","href":"/connections/github"},{"label":"Ambience vs Zep","href":"/vs/zep"}],"searchText":"Ambience for Cursor\nA governed team memory layer for Cursor and other MCP-speaking developer agents.\nAmbience gives Cursor access to scoped shared context through MCP, so project decisions and team conventions can follow the work rather than stay trapped in one user's local memory.\nMCP-native memory\nCursor can connect to the Ambience MCP server to search, read, and save memories. Ambience handles scope checks and redaction at the memory layer.\nProject context that survives\nWhen a Cursor session uncovers a durable project decision or failure mode, saving it to Ambience makes it available to future approved agents, including Claude Code and Codex.\nThe advantage over local-only memory\nLocal memory helps one developer. Ambience helps the team: shared scopes, source links, admin controls, conflict review, and audit history.\nWhat is governed agent memory?\nGitHub connection\nAmbience vs Zep"},{"type":"agent-runtime","title":"Ambience for GitHub Copilot","description":"Team context for GitHub Copilot workflows where useful coding decisions should become reusable memory.","path":"/agents/github-copilot","url":"https://ambience.sh/agents/github-copilot","markdownPath":"/agents/github-copilot.md","updated":"2026-06-07","summary":"Ambience helps Copilot-powered teams preserve decisions, conventions, and failure modes as governed context that can be reused across future agent sessions.","sections":[{"title":"Capture the reasoning around the code","body":["Copilot can help produce and review code, but the reasoning around a fix often disappears after the session. Ambience captures the durable parts as typed memories."]},{"title":"Works across agent surfaces","body":["The value is not limited to one IDE. Memories saved from Copilot workflows can inform Claude Code, Codex, Cursor, and MCP-speaking agents later, filtered by the user's access."]},{"title":"Governance stays central","body":["Ambience keeps source, scope, redaction, and audit attached to the memory so agent context can become team infrastructure."]}],"related":[{"label":"GitHub connection","href":"/connections/github"},{"label":"How do teams share AI agent memory?","href":"/answers/how-do-teams-share-ai-agent-memory"},{"label":"Install Ambience","href":"/install"}],"searchText":"Ambience for GitHub Copilot\nTeam context for GitHub Copilot workflows where useful coding decisions should become reusable memory.\nAmbience helps Copilot-powered teams preserve decisions, conventions, and failure modes as governed context that can be reused across future agent sessions.\nCapture the reasoning around the code\nCopilot can help produce and review code, but the reasoning around a fix often disappears after the session. Ambience captures the durable parts as typed memories.\nWorks across agent surfaces\nThe value is not limited to one IDE. Memories saved from Copilot workflows can inform Claude Code, Codex, Cursor, and MCP-speaking agents later, filtered by the user's access.\nGovernance stays central\nAmbience keeps source, scope, redaction, and audit attached to the memory so agent context can become team infrastructure.\nGitHub connection\nHow do teams share AI agent memory?\nInstall Ambience"},{"type":"company-context","title":"What is company context?","description":"A practical definition of company context for teams adopting AI agents at work.","path":"/company-context/what-is-company-context","url":"https://ambience.sh/company-context/what-is-company-context","markdownPath":"/company-context/what-is-company-context.md","updated":"2026-06-07","summary":"Company context is the reusable working memory of a business: the decisions, conventions, caveats, runbooks, sources, and permissions an AI agent needs before it can act safely.","proofPoints":["Company context is created by work, not by documentation alone.","It lives across calls, tickets, PRs, docs, Slack threads, and agent sessions.","Ambience turns that context into governed memory that approved agents can reuse."],"sections":[{"title":"The short definition","body":["Company context is the difference between a generic agent and an agent that understands how a specific organisation works.","It includes the why behind decisions, the practices teams actually follow, the sources that prove them, and the access rules that decide who should see them."]},{"title":"Why agents expose the gap","body":["Humans can ask around when context is missing. Agents usually cannot. If the relevant convention lives in a past call or a closed ticket, the agent starts from zero unless the company has a context layer.","Ambience gives that layer a durable shape: typed memory, scope, redaction, source, and audit."]},{"title":"The Ambience position","body":["A company context system should not be a passive archive. It should capture what changed, preserve where it came from, constrain who can use it, and make the next approved agent session start ahead."]}],"related":[{"label":"Company context for AI agents","href":"/company-context"},{"label":"Context Readiness Score","href":"/company-context/readiness-score"},{"label":"What is governed agent memory?","href":"/answers/what-is-governed-agent-memory"},{"label":"State of company context report","href":"/research/state-of-company-context-for-ai-agents-2026"}],"searchText":"What is company context?\nA practical definition of company context for teams adopting AI agents at work.\nCompany context is the reusable working memory of a business: the decisions, conventions, caveats, runbooks, sources, and permissions an AI agent needs before it can act safely.\nCompany context is created by work, not by documentation alone.\nIt lives across calls, tickets, PRs, docs, Slack threads, and agent sessions.\nAmbience turns that context into governed memory that approved agents can reuse.\nThe short definition\nCompany context is the difference between a generic agent and an agent that understands how a specific organisation works.\nIt includes the why behind decisions, the practices teams actually follow, the sources that prove them, and the access rules that decide who should see them.\nWhy agents expose the gap\nHumans can ask around when context is missing. Agents usually cannot. If the relevant convention lives in a past call or a closed ticket, the agent starts from zero unless the company has a context layer.\nAmbience gives that layer a durable shape: typed memory, scope, redaction, source, and audit.\nThe Ambience position\nA company context system should not be a passive archive. It should capture what changed, preserve where it came from, constrain who can use it, and make the next approved agent session start ahead.\nCompany context for AI agents\nContext Readiness Score\nWhat is governed agent memory?\nState of company context report"},{"type":"company-context","title":"Company Context Maturity Model","description":"A practical model for moving from scattered work history to governed company context for AI agents.","path":"/company-context/maturity-model","url":"https://ambience.sh/company-context/maturity-model","markdownPath":"/company-context/maturity-model.md","updated":"2026-06-07","summary":"Teams get leverage from agents when useful work becomes reusable context with source, scope, review, and reuse.","proofPoints":["The jump is from keeping records to changing the next agent run.","Each level adds a control agents need: source, scope, freshness, conflict review, and audit.","Ambience gives teams a path from first memories to context operations without forcing a large connector project first."],"sections":[{"title":"Level 0: cold starts","body":["Agents begin every session as if the company has no memory. Humans paste context, repeat decisions, and hope the agent notices the important parts.","This is where most teams start. It feels workable while usage is small, then breaks as soon as multiple people and agents touch the same work."]},{"title":"Level 1: personal recall","body":["One person has prompts, notes, local files, or private memory that helps their agent behave better. The context is useful, but it does not travel safely to teammates.","The risk is accidental authority. A private preference starts to look like a team rule because the agent has no better source."]},{"title":"Level 2: project memory","body":["Important decisions and conventions are attached to a project. Agents can retrieve the current implementation constraints before they edit, plan, or respond.","Ambience makes this level tangible with project-scoped memories, source links, redaction state, and session-start context."]},{"title":"Level 3: source-linked team memory","body":["Calls, tickets, PRs, docs, threads, and agent sessions become approved source material. The memory is the durable takeaway, not the whole artifact.","This is the first compounding level. A Granola call can inform a Codex session. A GitHub PR can inform a future Linear plan. The source remains inspectable."]},{"title":"Level 4: governed company context","body":["Context is shared across the organisation with scopes, redaction, conflict review, access changes, and audit. Agents receive context because they are allowed to see it, not only because it matches a query.","This is where Ambience becomes infrastructure: memory reads and writes are visible product events rather than invisible prompt text."]},{"title":"Level 5: context operations","body":["Teams run lightweight context ops: weekly reviews, source sweeps, stale-memory cleanup, skill promotion, and context readiness checks before large agent rollouts.","The outcome is not a perfect archive. It is a living company context map that keeps agents close to the work the business actually trusts."]}],"related":[{"label":"Context Readiness Score","href":"/company-context/readiness-score"},{"label":"Company Context Map","href":"/company-context/company-context-map"},{"label":"How to build company context with Ambience","href":"/blog/build-company-context-with-ambience"}],"searchText":"Company Context Maturity Model\nA practical model for moving from scattered work history to governed company context for AI agents.\nTeams get leverage from agents when useful work becomes reusable context with source, scope, review, and reuse.\nThe jump is from keeping records to changing the next agent run.\nEach level adds a control agents need: source, scope, freshness, conflict review, and audit.\nAmbience gives teams a path from first memories to context operations without forcing a large connector project first.\nLevel 0: cold starts\nAgents begin every session as if the company has no memory. Humans paste context, repeat decisions, and hope the agent notices the important parts.\nThis is where most teams start. It feels workable while usage is small, then breaks as soon as multiple people and agents touch the same work.\nLevel 1: personal recall\nOne person has prompts, notes, local files, or private memory that helps their agent behave better. The context is useful, but it does not travel safely to teammates.\nThe risk is accidental authority. A private preference starts to look like a team rule because the agent has no better source.\nLevel 2: project memory\nImportant decisions and conventions are attached to a project. Agents can retrieve the current implementation constraints before they edit, plan, or respond.\nAmbience makes this level tangible with project-scoped memories, source links, redaction state, and session-start context.\nLevel 3: source-linked team memory\nCalls, tickets, PRs, docs, threads, and agent sessions become approved source material. The memory is the durable takeaway, not the whole artifact.\nThis is the first compounding level. A Granola call can inform a Codex session. A GitHub PR can inform a future Linear plan. The source remains inspectable.\nLevel 4: governed company context\nContext is shared across the organisation with scopes, redaction, conflict review, access changes, and audit. Agents receive context because they are allowed to see it, not only because it matches a query.\nThis is where Ambience becomes infrastructure: memory reads and writes are visible product events rather than invisible prompt text.\nLevel 5: context operations\nTeams run lightweight context ops: weekly reviews, source sweeps, stale-memory cleanup, skill promotion, and context readiness checks before large agent rollouts.\nThe outcome is not a perfect archive. It is a living company context map that keeps agents close to the work the business actually trusts.\nContext Readiness Score\nCompany Context Map\nHow to build company context with Ambience"},{"type":"company-context","title":"Context Readiness Score","description":"A short Ambience audit for whether a team is ready to let agents rely on company context.","path":"/company-context/readiness-score","url":"https://ambience.sh/company-context/readiness-score","markdownPath":"/company-context/readiness-score.md","updated":"2026-06-07","summary":"Readiness is not about how much information the company has. It is whether the important context has source, scope, redaction, review, and a route into the next agent session.","proofPoints":["The score checks operational controls, not abstract AI maturity.","It turns onboarding into a permissioned sweep, review, and first seeded context set.","It gives teams a simple next action: capture sources, narrow scope, review conflicts, or teach agents the Ambience loop."],"sections":[{"title":"What the score tells you","body":["A high score means agents can start with selected, current, permissioned context. A low score means the company still depends on people manually restating decisions.","The useful question is not whether context exists. It does. The useful question is whether an approved agent can retrieve the right memory, trust its source, and leave evidence behind."]},{"title":"How to use the audit","body":["Run it before onboarding a new team, before turning on regular sweeps, and after any major change in agent access. Keep the answers evidence-based.","If an answer is uncertain, treat it as no for now. Ambience makes proof visible enough that the team does not have to guess."]},{"title":"What Ambience improves first","body":["Start with one active project and one week of approved sources. Ask the agent to propose candidate decisions, conventions, failures, skills, patterns, and references. Review them with source, scope, and redaction visible before anything becomes shared memory."]}],"related":[{"label":"Company Context Maturity Model","href":"/company-context/maturity-model"},{"label":"Onboarding new agents","href":"/company-context/onboarding-new-agents"},{"label":"How to audit AI agent context","href":"/answers/how-to-audit-ai-agent-context"}],"searchText":"Context Readiness Score\nA short Ambience audit for whether a team is ready to let agents rely on company context.\nReadiness is not about how much information the company has. It is whether the important context has source, scope, redaction, review, and a route into the next agent session.\nThe score checks operational controls, not abstract AI maturity.\nIt turns onboarding into a permissioned sweep, review, and first seeded context set.\nIt gives teams a simple next action: capture sources, narrow scope, review conflicts, or teach agents the Ambience loop.\nWhat the score tells you\nA high score means agents can start with selected, current, permissioned context. A low score means the company still depends on people manually restating decisions.\nThe useful question is not whether context exists. It does. The useful question is whether an approved agent can retrieve the right memory, trust its source, and leave evidence behind.\nHow to use the audit\nRun it before onboarding a new team, before turning on regular sweeps, and after any major change in agent access. Keep the answers evidence-based.\nIf an answer is uncertain, treat it as no for now. Ambience makes proof visible enough that the team does not have to guess.\nWhat Ambience improves first\nStart with one active project and one week of approved sources. Ask the agent to propose candidate decisions, conventions, failures, skills, patterns, and references. Review them with source, scope, and redaction visible before anything becomes shared memory.\nCompany Context Maturity Model\nOnboarding new agents\nHow to audit AI agent context"},{"type":"company-context","title":"Company Context Map","description":"The Ambience map of sources, memories, scopes, freshness, conflicts, and agent runs.","path":"/company-context/company-context-map","url":"https://ambience.sh/company-context/company-context-map","markdownPath":"/company-context/company-context-map.md","updated":"2026-06-07","summary":"A company context map lets a team see what agents know, where it came from, who can use it, and what needs review.","proofPoints":["The map starts with sources agents already reach: calls, tickets, PRs, docs, threads, files, and sessions.","It separates source artifacts from durable Ambience memories so teams do not save noise.","It makes context health visible through freshness, conflict state, scope, and audit."],"sections":[{"title":"Sources","body":["Sources are the work artifacts where context is born: a Granola call, Linear ticket, GitHub PR, Slack thread, Google Doc, Notion page, local repository, or agent session.","Ambience does not need every direct connector before it can help. If an approved agent can already read a source, Ambience can ask permission and propose source-linked memories from it."]},{"title":"Memories","body":["Memories are the selected takeaways future agents should reuse: decisions, conventions, patterns, skills, failures, and references.","The memory is smaller than the source and more accountable than a summary. It names the durable claim and links back to evidence."]},{"title":"Scopes","body":["Scopes decide who can use the memory. Personal, team, project, org, and sensitive scopes keep retrieval aligned with permissions instead of pure relevance.","The context map makes over-broad scopes easy to spot before a memory spreads into the wrong work."]},{"title":"Freshness and conflicts","body":["Context goes stale. Ambience treats conflicts as a product surface: review, keep both, accept a newer memory, or dismiss a false conflict with evidence.","Freshness does not mean deleting history. It means making the current instruction clear enough that agents do not improvise."]},{"title":"Agent runs","body":["The map closes the loop by showing which agent runs read context, wrote memories, or changed access. This is what makes shared memory auditable.","A team can ask why an agent acted a certain way and see the context packet that shaped the answer."]}],"related":[{"label":"Source-linked decision example","href":"/examples/source-linked-decision"},{"label":"Security and permissions","href":"/company-context/security-and-permissions"},{"label":"Context Readiness Score","href":"/company-context/readiness-score"}],"searchText":"Company Context Map\nThe Ambience map of sources, memories, scopes, freshness, conflicts, and agent runs.\nA company context map lets a team see what agents know, where it came from, who can use it, and what needs review.\nThe map starts with sources agents already reach: calls, tickets, PRs, docs, threads, files, and sessions.\nIt separates source artifacts from durable Ambience memories so teams do not save noise.\nIt makes context health visible through freshness, conflict state, scope, and audit.\nSources\nSources are the work artifacts where context is born: a Granola call, Linear ticket, GitHub PR, Slack thread, Google Doc, Notion page, local repository, or agent session.\nAmbience does not need every direct connector before it can help. If an approved agent can already read a source, Ambience can ask permission and propose source-linked memories from it.\nMemories\nMemories are the selected takeaways future agents should reuse: decisions, conventions, patterns, skills, failures, and references.\nThe memory is smaller than the source and more accountable than a summary. It names the durable claim and links back to evidence.\nScopes\nScopes decide who can use the memory. Personal, team, project, org, and sensitive scopes keep retrieval aligned with permissions instead of pure relevance.\nThe context map makes over-broad scopes easy to spot before a memory spreads into the wrong work.\nFreshness and conflicts\nContext goes stale. Ambience treats conflicts as a product surface: review, keep both, accept a newer memory, or dismiss a false conflict with evidence.\nFreshness does not mean deleting history. It means making the current instruction clear enough that agents do not improvise.\nAgent runs\nThe map closes the loop by showing which agent runs read context, wrote memories, or changed access. This is what makes shared memory auditable.\nA team can ask why an agent acted a certain way and see the context packet that shaped the answer.\nSource-linked decision example\nSecurity and permissions\nContext Readiness Score"},{"type":"company-context","title":"Agent brief for company context","description":"A concise brief for agents using Ambience as the governed company context layer.","path":"/company-context/agent-brief","url":"https://ambience.sh/company-context/agent-brief","markdownPath":"/company-context/agent-brief.md","updated":"2026-06-07","summary":"Agents should read Ambience context before meaningful work, search it when decisions are unclear, and save only durable memories with source, scope, type, and redaction state.","proofPoints":["Do not save raw transcripts or broad summaries when a typed memory is enough.","Use source links as evidence, not decoration.","Default to the narrowest scope that still helps the next approved agent."],"sections":[{"title":"Before work","body":["Load the session-start memories Ambience provides. Search Ambience for active project decisions, team conventions, failures, skills, and references before making a plan that depends on company context.","If the context is missing, say so. Do not invent a team decision because a document sounded plausible."]},{"title":"During work","body":["When the user refers to a call, ticket, PR, document, thread, local file, or connected MCP, treat it as possible source material. Save the durable takeaway only when it will change future work.","Keep the candidate memory concise. Name the decision, convention, pattern, failure, skill, or reference and the reason it matters."]},{"title":"After work","body":["Save durable decisions when they land. Attach source and scope. Redact secrets and unnecessary personal details before storage.","If a new memory conflicts with an old one, surface the conflict for review instead of quietly overwriting the organisational record."]}],"related":[{"label":"Full agent reference","href":"/llms-full.txt"},{"label":"How to build company context with Ambience","href":"/blog/build-company-context-with-ambience"},{"label":"MCP memory","href":"/glossary/mcp-memory"}],"searchText":"Agent brief for company context\nA concise brief for agents using Ambience as the governed company context layer.\nAgents should read Ambience context before meaningful work, search it when decisions are unclear, and save only durable memories with source, scope, type, and redaction state.\nDo not save raw transcripts or broad summaries when a typed memory is enough.\nUse source links as evidence, not decoration.\nDefault to the narrowest scope that still helps the next approved agent.\nBefore work\nLoad the session-start memories Ambience provides. Search Ambience for active project decisions, team conventions, failures, skills, and references before making a plan that depends on company context.\nIf the context is missing, say so. Do not invent a team decision because a document sounded plausible.\nDuring work\nWhen the user refers to a call, ticket, PR, document, thread, local file, or connected MCP, treat it as possible source material. Save the durable takeaway only when it will change future work.\nKeep the candidate memory concise. Name the decision, convention, pattern, failure, skill, or reference and the reason it matters.\nAfter work\nSave durable decisions when they land. Attach source and scope. Redact secrets and unnecessary personal details before storage.\nIf a new memory conflicts with an old one, surface the conflict for review instead of quietly overwriting the organisational record.\nFull agent reference\nHow to build company context with Ambience\nMCP memory"},{"type":"company-context","title":"Company context vs knowledge base","description":"Why company context for agents is different from a knowledge base built for humans.","path":"/company-context/company-context-vs-knowledge-base","url":"https://ambience.sh/company-context/company-context-vs-knowledge-base","markdownPath":"/company-context/company-context-vs-knowledge-base.md","updated":"2026-06-07","summary":"A knowledge base helps humans look up information. Company context helps approved agents act with the right decisions, constraints, and source evidence already loaded.","proofPoints":["Knowledge bases are broad and human-readable.","Company context is selective, scoped, and action-oriented.","Ambience stores the durable agent-facing takeaway and keeps the source visible."],"sections":[{"title":"The archive problem","body":["A knowledge base can contain everything and still fail an agent. The agent needs the few pieces of context that change what it should do right now.","That means decisions, conventions, failures, and references need to be extracted from broad documents into durable memory."]},{"title":"The governance problem","body":["Human search tools often assume the reader can judge what applies. Agents need stronger boundaries. They need scope checks, source links, and stale-context handling before retrieval reaches the prompt."]},{"title":"Why Ambience is different","body":["Ambience does not replace Notion, Confluence, Google Drive, Slack, or GitHub. It turns the durable parts of those systems into governed memory that agents can reuse."]}],"related":[{"label":"Ambience vs Confluence","href":"/compare/confluence"},{"label":"Ambience vs Notion AI","href":"/compare/notion-ai"},{"label":"Google Docs connection","href":"/connections/google-docs"}],"searchText":"Company context vs knowledge base\nWhy company context for agents is different from a knowledge base built for humans.\nA knowledge base helps humans look up information. Company context helps approved agents act with the right decisions, constraints, and source evidence already loaded.\nKnowledge bases are broad and human-readable.\nCompany context is selective, scoped, and action-oriented.\nAmbience stores the durable agent-facing takeaway and keeps the source visible.\nThe archive problem\nA knowledge base can contain everything and still fail an agent. The agent needs the few pieces of context that change what it should do right now.\nThat means decisions, conventions, failures, and references need to be extracted from broad documents into durable memory.\nThe governance problem\nHuman search tools often assume the reader can judge what applies. Agents need stronger boundaries. They need scope checks, source links, and stale-context handling before retrieval reaches the prompt.\nWhy Ambience is different\nAmbience does not replace Notion, Confluence, Google Drive, Slack, or GitHub. It turns the durable parts of those systems into governed memory that agents can reuse.\nAmbience vs Confluence\nAmbience vs Notion AI\nGoogle Docs connection"},{"type":"company-context","title":"Company context vs agent memory","description":"The category split between single-agent recall and governed company context.","path":"/company-context/company-context-vs-agent-memory","url":"https://ambience.sh/company-context/company-context-vs-agent-memory","markdownPath":"/company-context/company-context-vs-agent-memory.md","updated":"2026-06-07","summary":"Agent memory usually means one agent remembering past interactions. Company context means a whole organisation sharing trusted context across people, projects, and agents.","proofPoints":["Single-agent memory optimizes recall.","Company context optimizes safe reuse across a team.","Ambience is built for the team layer: scopes, redaction, source links, conflict review, and audit."],"sections":[{"title":"Different failure modes","body":["A personal memory system fails when it forgets. A company context system fails when it shows the wrong context to the wrong person, repeats a stale decision, or cannot prove why an agent acted."]},{"title":"Different architecture","body":["Personal memory can be greedy. Company context must be conservative. It needs permission checks and source evidence before relevance ranking is enough."]},{"title":"Ambience's category","body":["Ambience is not trying to be the best private memory for one user. It is the governed context plane that makes many agents across a company start smarter."]}],"related":[{"label":"Single-agent memory vs team context","href":"/writing/agent-memory-vs-team-context"},{"label":"Ambience vs Mem0","href":"/vs/mem0"},{"label":"How do teams share AI agent memory?","href":"/answers/how-do-teams-share-ai-agent-memory"}],"searchText":"Company context vs agent memory\nThe category split between single-agent recall and governed company context.\nAgent memory usually means one agent remembering past interactions. Company context means a whole organisation sharing trusted context across people, projects, and agents.\nSingle-agent memory optimizes recall.\nCompany context optimizes safe reuse across a team.\nAmbience is built for the team layer: scopes, redaction, source links, conflict review, and audit.\nDifferent failure modes\nA personal memory system fails when it forgets. A company context system fails when it shows the wrong context to the wrong person, repeats a stale decision, or cannot prove why an agent acted.\nDifferent architecture\nPersonal memory can be greedy. Company context must be conservative. It needs permission checks and source evidence before relevance ranking is enough.\nAmbience's category\nAmbience is not trying to be the best private memory for one user. It is the governed context plane that makes many agents across a company start smarter.\nSingle-agent memory vs team context\nAmbience vs Mem0\nHow do teams share AI agent memory?"},{"type":"company-context","title":"Company context for AI agents","description":"What AI agents need from company context before they can work reliably across a team.","path":"/company-context/for-ai-agents","url":"https://ambience.sh/company-context/for-ai-agents","markdownPath":"/company-context/for-ai-agents.md","updated":"2026-06-07","summary":"AI agents need context that is current, source-linked, scoped, and short enough to act on. Ambience provides that company context layer.","proofPoints":["Agents need compact operational context, not full archives.","Source-linked memories let agents cite and inspect the origin of a decision.","Scopes and audit let teams trust context reuse."],"sections":[{"title":"The context an agent actually needs","body":["Before editing code, replying to a customer, or drafting a plan, an agent needs the team's current decisions, constraints, conventions, and known failure modes.","Those facts are usually scattered. Ambience gives them a typed, retrievable form."]},{"title":"The prompt is not the system","body":["Stuffing more text into a prompt is a temporary fix. Company context needs lifecycle: capture, redaction, scope, conflict review, retrieval, and audit."]},{"title":"The Ambience layer","body":["Ambience sits between approved agents and the memory corpus. It decides what can be read, what can be saved, and what source evidence travels with the memory."]}],"related":[{"label":"Ambience for Claude Code","href":"/agents/claude-code"},{"label":"Ambience for Codex","href":"/agents/codex"},{"label":"How to stop agents starting from zero","href":"/answers/how-to-stop-agents-starting-from-zero"}],"searchText":"Company context for AI agents\nWhat AI agents need from company context before they can work reliably across a team.\nAI agents need context that is current, source-linked, scoped, and short enough to act on. Ambience provides that company context layer.\nAgents need compact operational context, not full archives.\nSource-linked memories let agents cite and inspect the origin of a decision.\nScopes and audit let teams trust context reuse.\nThe context an agent actually needs\nBefore editing code, replying to a customer, or drafting a plan, an agent needs the team's current decisions, constraints, conventions, and known failure modes.\nThose facts are usually scattered. Ambience gives them a typed, retrievable form.\nThe prompt is not the system\nStuffing more text into a prompt is a temporary fix. Company context needs lifecycle: capture, redaction, scope, conflict review, retrieval, and audit.\nThe Ambience layer\nAmbience sits between approved agents and the memory corpus. It decides what can be read, what can be saved, and what source evidence travels with the memory.\nAmbience for Claude Code\nAmbience for Codex\nHow to stop agents starting from zero"},{"type":"company-context","title":"Company context for Claude Code","description":"How Claude Code teams can use Ambience as the company context layer for engineering agents.","path":"/company-context/for-claude-code","url":"https://ambience.sh/company-context/for-claude-code","markdownPath":"/company-context/for-claude-code.md","updated":"2026-06-07","summary":"Ambience gives Claude Code the company context it needs at session start and captures durable decisions at session end.","proofPoints":["Project decisions and conventions can follow the repository.","A teammate's prior Claude Code session can help the next teammate without sharing raw transcripts.","Audit and revocation make shared context manageable for teams."],"sections":[{"title":"Why Claude Code needs company context","body":["Claude Code can read a repository, but it cannot infer every decision the team made in Slack, Linear, GitHub, Granola, or prior sessions.","Ambience gives Claude Code a trusted starting packet of relevant company context."]},{"title":"What Ambience captures","body":["The durable parts: decisions, conventions, implementation patterns, runbooks, failures, and references. The source stays visible and the memory is scoped before reuse."]}],"related":[{"label":"Ambience for Claude Code","href":"/agents/claude-code"},{"label":"GitHub PR to engineering memory","href":"/examples/github-pr-to-engineering-memory"},{"label":"Best memory layer for Claude Code teams","href":"/answers/best-memory-layer-for-claude-code-teams"}],"searchText":"Company context for Claude Code\nHow Claude Code teams can use Ambience as the company context layer for engineering agents.\nAmbience gives Claude Code the company context it needs at session start and captures durable decisions at session end.\nProject decisions and conventions can follow the repository.\nA teammate's prior Claude Code session can help the next teammate without sharing raw transcripts.\nAudit and revocation make shared context manageable for teams.\nWhy Claude Code needs company context\nClaude Code can read a repository, but it cannot infer every decision the team made in Slack, Linear, GitHub, Granola, or prior sessions.\nAmbience gives Claude Code a trusted starting packet of relevant company context.\nWhat Ambience captures\nThe durable parts: decisions, conventions, implementation patterns, runbooks, failures, and references. The source stays visible and the memory is scoped before reuse.\nAmbience for Claude Code\nGitHub PR to engineering memory\nBest memory layer for Claude Code teams"},{"type":"company-context","title":"Company context for Codex","description":"How Codex can use Ambience to start with trusted team context instead of rediscovering work.","path":"/company-context/for-codex","url":"https://ambience.sh/company-context/for-codex","markdownPath":"/company-context/for-codex.md","updated":"2026-06-07","summary":"Ambience gives Codex scoped company context through memory summaries, MCP, and CLI workflows.","proofPoints":["Codex can use Ambience memory to understand project decisions and team conventions.","Durable saves make future sessions smarter.","MCP and CLI parity keeps the workflow agent-native."],"sections":[{"title":"The Codex loop","body":["A Codex session starts with relevant memories, performs the work, and saves durable decisions when they land. Future agents inherit the result instead of repeating the investigation."]},{"title":"What makes it company context","body":["The memory is not private prompt history. It carries source, scope, redaction state, and audit events so teams can trust it."]}],"related":[{"label":"Ambience for Codex","href":"/agents/codex"},{"label":"How to stop agents starting from zero","href":"/answers/how-to-stop-agents-starting-from-zero"},{"label":"Linear ticket to project context","href":"/examples/linear-ticket-to-project-context"}],"searchText":"Company context for Codex\nHow Codex can use Ambience to start with trusted team context instead of rediscovering work.\nAmbience gives Codex scoped company context through memory summaries, MCP, and CLI workflows.\nCodex can use Ambience memory to understand project decisions and team conventions.\nDurable saves make future sessions smarter.\nMCP and CLI parity keeps the workflow agent-native.\nThe Codex loop\nA Codex session starts with relevant memories, performs the work, and saves durable decisions when they land. Future agents inherit the result instead of repeating the investigation.\nWhat makes it company context\nThe memory is not private prompt history. It carries source, scope, redaction state, and audit events so teams can trust it.\nAmbience for Codex\nHow to stop agents starting from zero\nLinear ticket to project context"},{"type":"company-context","title":"Company context from Slack","description":"How important Slack decisions can become governed company context for AI agents.","path":"/company-context/from-slack","url":"https://ambience.sh/company-context/from-slack","markdownPath":"/company-context/from-slack.md","updated":"2026-06-07","summary":"Slack contains decisions agents need later. Ambience lets approved agents preserve the durable decision as scoped memory with Slack as the source.","proofPoints":["The memory is the decision, not the whole thread.","The source remains inspectable.","The user chooses what becomes shared context."],"sections":[{"title":"The Slack failure mode","body":["Teams decide things in Slack and then lose them to scrollback. Agents feel that loss immediately because they cannot know which thread mattered."]},{"title":"The Ambience pattern","body":["An approved agent reads the selected thread, proposes durable memories, and saves the confirmed context to Ambience with scope and source attached."]}],"related":[{"label":"Slack connection","href":"/connections/slack"},{"label":"Slack thread to team convention","href":"/examples/slack-thread-to-team-convention"},{"label":"Security and trust","href":"/security"}],"searchText":"Company context from Slack\nHow important Slack decisions can become governed company context for AI agents.\nSlack contains decisions agents need later. Ambience lets approved agents preserve the durable decision as scoped memory with Slack as the source.\nThe memory is the decision, not the whole thread.\nThe source remains inspectable.\nThe user chooses what becomes shared context.\nThe Slack failure mode\nTeams decide things in Slack and then lose them to scrollback. Agents feel that loss immediately because they cannot know which thread mattered.\nThe Ambience pattern\nAn approved agent reads the selected thread, proposes durable memories, and saves the confirmed context to Ambience with scope and source attached.\nSlack connection\nSlack thread to team convention\nSecurity and trust"},{"type":"company-context","title":"Company context from Linear","description":"How Linear tickets and project decisions become reusable company context in Ambience.","path":"/company-context/from-linear","url":"https://ambience.sh/company-context/from-linear","markdownPath":"/company-context/from-linear.md","updated":"2026-06-07","summary":"Linear captures what the team plans to build. Ambience captures the decisions and caveats future agents need while building it.","proofPoints":["Tickets often contain the reason behind a project decision.","Project-scoped memories can follow the implementation work.","Future agents can reuse Linear-derived context without reading every issue."],"sections":[{"title":"The planning-to-memory path","body":["Approved Linear issues can seed decisions, risks, references, and failure memories. Ambience keeps Linear as the source and stores the durable context agents should reuse."]},{"title":"Why scope matters","body":["A project-specific caveat should not become an org-wide rule by accident. Ambience gives Linear-derived context a scope before future agents see it."]}],"related":[{"label":"Linear connection","href":"/connections/linear"},{"label":"Linear ticket to project context","href":"/examples/linear-ticket-to-project-context"},{"label":"How scoped memory works for teams","href":"/writing/scoped-memory-for-teams"}],"searchText":"Company context from Linear\nHow Linear tickets and project decisions become reusable company context in Ambience.\nLinear captures what the team plans to build. Ambience captures the decisions and caveats future agents need while building it.\nTickets often contain the reason behind a project decision.\nProject-scoped memories can follow the implementation work.\nFuture agents can reuse Linear-derived context without reading every issue.\nThe planning-to-memory path\nApproved Linear issues can seed decisions, risks, references, and failure memories. Ambience keeps Linear as the source and stores the durable context agents should reuse.\nWhy scope matters\nA project-specific caveat should not become an org-wide rule by accident. Ambience gives Linear-derived context a scope before future agents see it.\nLinear connection\nLinear ticket to project context\nHow scoped memory works for teams"},{"type":"company-context","title":"Company context from Granola","description":"How meeting decisions from Granola become source-linked company context for agents.","path":"/company-context/from-granola","url":"https://ambience.sh/company-context/from-granola","markdownPath":"/company-context/from-granola.md","updated":"2026-06-07","summary":"Granola captures the call. Ambience captures the durable decision, owner, constraint, or reference that future agents need.","proofPoints":["Meeting notes are a source, not the final memory shape.","Ambience stores the reusable takeaway with the call as provenance.","Redaction and scope happen before the decision becomes team context."],"sections":[{"title":"Meetings create context agents need","body":["A customer caveat, onboarding decision, or rollout constraint may first appear in a call. If it stays only in meeting notes, future agents start without it."]},{"title":"The Ambience pattern","body":["An approved agent proposes memories from the selected call. The user confirms what should be saved, and Ambience stores the redacted memory with Granola as source."]}],"related":[{"label":"Granola connection","href":"/connections/granola"},{"label":"Granola call to memory","href":"/examples/granola-call-to-memory"},{"label":"How to audit AI agent context","href":"/answers/how-to-audit-ai-agent-context"}],"searchText":"Company context from Granola\nHow meeting decisions from Granola become source-linked company context for agents.\nGranola captures the call. Ambience captures the durable decision, owner, constraint, or reference that future agents need.\nMeeting notes are a source, not the final memory shape.\nAmbience stores the reusable takeaway with the call as provenance.\nRedaction and scope happen before the decision becomes team context.\nMeetings create context agents need\nA customer caveat, onboarding decision, or rollout constraint may first appear in a call. If it stays only in meeting notes, future agents start without it.\nThe Ambience pattern\nAn approved agent proposes memories from the selected call. The user confirms what should be saved, and Ambience stores the redacted memory with Granola as source.\nGranola connection\nGranola call to memory\nHow to audit AI agent context"},{"type":"company-context","title":"Security and permissions for company context","description":"The controls required before company context can safely power AI agents.","path":"/company-context/security-and-permissions","url":"https://ambience.sh/company-context/security-and-permissions","markdownPath":"/company-context/security-and-permissions.md","updated":"2026-06-07","summary":"Company context becomes risky when it lacks source, scope, redaction, revocation, conflict review, and audit. Ambience makes those controls part of the memory layer.","proofPoints":["Agents should not receive context only because it is relevant.","Raw secrets should not become durable memory.","Admins need proof of who read which context and why."],"sections":[{"title":"Relevance is not permission","body":["A memory can match a query and still be inappropriate for the current user, project, or agent. Company context needs access control before retrieval reaches the prompt."]},{"title":"The trust controls","body":["Ambience uses redaction before storage, five scopes, source-linked provenance, revocation, conflict review, and append-only audit."]}],"related":[{"label":"Security and trust","href":"/security"},{"label":"Why redaction has to happen before storage","href":"/writing/redaction-before-storage"},{"label":"How to audit AI agent context","href":"/answers/how-to-audit-ai-agent-context"}],"searchText":"Security and permissions for company context\nThe controls required before company context can safely power AI agents.\nCompany context becomes risky when it lacks source, scope, redaction, revocation, conflict review, and audit. Ambience makes those controls part of the memory layer.\nAgents should not receive context only because it is relevant.\nRaw secrets should not become durable memory.\nAdmins need proof of who read which context and why.\nRelevance is not permission\nA memory can match a query and still be inappropriate for the current user, project, or agent. Company context needs access control before retrieval reaches the prompt.\nThe trust controls\nAmbience uses redaction before storage, five scopes, source-linked provenance, revocation, conflict review, and append-only audit.\nSecurity and trust\nWhy redaction has to happen before storage\nHow to audit AI agent context"},{"type":"company-context","title":"Onboarding new agents with company context","description":"How teams can make new agent sessions inherit the context the organisation already earned.","path":"/company-context/onboarding-new-agents","url":"https://ambience.sh/company-context/onboarding-new-agents","markdownPath":"/company-context/onboarding-new-agents.md","updated":"2026-06-07","summary":"New agents do not start from zero. Ambience can seed them with scoped decisions, conventions, failures, skills, and references from approved sources.","proofPoints":["Onboarding asks which sources are included.","Seeded memories are reviewed before becoming shared context.","Future sessions show evidence that the context was reused."],"sections":[{"title":"The first-session problem","body":["New agents do not know the team's conventions, recent decisions, or known traps. Human teammates spend time re-explaining context the organisation already earned."]},{"title":"The Ambience onboarding loop","body":["Ambience can discover available agent connections, ask what the user wants included, propose memories from approved sources, and seed future sessions with the confirmed context."]}],"related":[{"label":"How to stop agents starting from zero","href":"/answers/how-to-stop-agents-starting-from-zero"},{"label":"Ambience for Claude Code","href":"/agents/claude-code"},{"label":"Install Ambience","href":"/install"}],"searchText":"Onboarding new agents with company context\nHow teams can make new agent sessions inherit the context the organisation already earned.\nNew agents do not start from zero. Ambience can seed them with scoped decisions, conventions, failures, skills, and references from approved sources.\nOnboarding asks which sources are included.\nSeeded memories are reviewed before becoming shared context.\nFuture sessions show evidence that the context was reused.\nThe first-session problem\nNew agents do not know the team's conventions, recent decisions, or known traps. Human teammates spend time re-explaining context the organisation already earned.\nThe Ambience onboarding loop\nAmbience can discover available agent connections, ask what the user wants included, propose memories from approved sources, and seed future sessions with the confirmed context.\nHow to stop agents starting from zero\nAmbience for Claude Code\nInstall Ambience"},{"type":"example","title":"Example: source-linked decision memory","description":"A concrete example of an Ambience decision memory with source, scope, redaction, and future retrieval.","path":"/examples/source-linked-decision","url":"https://ambience.sh/examples/source-linked-decision","markdownPath":"/examples/source-linked-decision.md","updated":"2026-06-07","summary":"A source-linked decision memory turns a messy work artifact into a concise, auditable record an agent can reuse.","proofPoints":["Source: the original call, ticket, PR, thread, document, or session.","Memory: the reusable decision and reasoning.","Control: scope, redaction, conflict status, and audit."],"sections":[{"title":"Source","body":["Granola call: the team agreed that onboarding sweeps should ask users which connected apps are included before scanning any source material."]},{"title":"Ambience memory","body":["Decision: Ask users to approve included sources before onboarding sweeps. Scope: org. Source: Granola call. Reason: it keeps setup permissioned and prevents accidental import of private context."]},{"title":"Future retrieval","body":["When an agent works on onboarding, Ambience can retrieve the decision and show where it came from before the agent proposes a flow."]}],"related":[{"label":"Granola call to memory","href":"/examples/granola-call-to-memory"},{"label":"How to audit AI agent context","href":"/answers/how-to-audit-ai-agent-context"},{"label":"Company context","href":"/company-context"}],"searchText":"Example: source-linked decision memory\nA concrete example of an Ambience decision memory with source, scope, redaction, and future retrieval.\nA source-linked decision memory turns a messy work artifact into a concise, auditable record an agent can reuse.\nSource: the original call, ticket, PR, thread, document, or session.\nMemory: the reusable decision and reasoning.\nControl: scope, redaction, conflict status, and audit.\nSource\nGranola call: the team agreed that onboarding sweeps should ask users which connected apps are included before scanning any source material.\nAmbience memory\nDecision: Ask users to approve included sources before onboarding sweeps. Scope: org. Source: Granola call. Reason: it keeps setup permissioned and prevents accidental import of private context.\nFuture retrieval\nWhen an agent works on onboarding, Ambience can retrieve the decision and show where it came from before the agent proposes a flow.\nGranola call to memory\nHow to audit AI agent context\nCompany context"},{"type":"example","title":"Example: Granola call to Ambience memory","description":"How a meeting note becomes a source-linked Ambience memory for future agents.","path":"/examples/granola-call-to-memory","url":"https://ambience.sh/examples/granola-call-to-memory","markdownPath":"/examples/granola-call-to-memory.md","updated":"2026-06-07","summary":"The call stays in Granola. Ambience stores the durable decision, owner, and caveat future agents need.","proofPoints":["Meeting notes are noisy.","The useful artifact is a confirmed decision memory.","The source link lets the team inspect the original call when needed."],"sections":[{"title":"Before Ambience","body":["A customer onboarding call contains a decision about which integrations are safe to sweep. The note is accurate, but no future coding agent will read it by default."]},{"title":"After Ambience","body":["The agent proposes a decision memory from the call. The user confirms it, Ambience redacts sensitive details, and the memory becomes project or org context."]}],"related":[{"label":"Granola connection","href":"/connections/granola"},{"label":"Company context from Granola","href":"/company-context/from-granola"},{"label":"Source-linked decision","href":"/examples/source-linked-decision"}],"searchText":"Example: Granola call to Ambience memory\nHow a meeting note becomes a source-linked Ambience memory for future agents.\nThe call stays in Granola. Ambience stores the durable decision, owner, and caveat future agents need.\nMeeting notes are noisy.\nThe useful artifact is a confirmed decision memory.\nThe source link lets the team inspect the original call when needed.\nBefore Ambience\nA customer onboarding call contains a decision about which integrations are safe to sweep. The note is accurate, but no future coding agent will read it by default.\nAfter Ambience\nThe agent proposes a decision memory from the call. The user confirms it, Ambience redacts sensitive details, and the memory becomes project or org context.\nGranola connection\nCompany context from Granola\nSource-linked decision"},{"type":"example","title":"Example: Linear ticket to project context","description":"How a Linear issue becomes scoped Ambience project memory.","path":"/examples/linear-ticket-to-project-context","url":"https://ambience.sh/examples/linear-ticket-to-project-context","markdownPath":"/examples/linear-ticket-to-project-context.md","updated":"2026-06-07","summary":"A Linear ticket can produce a project-scoped decision memory that future engineering agents retrieve before implementation.","proofPoints":["The ticket remains the source of record.","Ambience stores the implementation constraint.","Project scope prevents accidental org-wide generalization."],"sections":[{"title":"Source","body":["Linear issue: use WorkOS org membership as the access boundary for the new connections surface."]},{"title":"Memory","body":["Decision: Use WorkOS org membership as the access boundary. Scope: project. Source: Linear issue. Future agents should preserve this unless an admin access decision supersedes it."]}],"related":[{"label":"Linear connection","href":"/connections/linear"},{"label":"Company context from Linear","href":"/company-context/from-linear"},{"label":"How scoped memory works for teams","href":"/writing/scoped-memory-for-teams"}],"searchText":"Example: Linear ticket to project context\nHow a Linear issue becomes scoped Ambience project memory.\nA Linear ticket can produce a project-scoped decision memory that future engineering agents retrieve before implementation.\nThe ticket remains the source of record.\nAmbience stores the implementation constraint.\nProject scope prevents accidental org-wide generalization.\nSource\nLinear issue: use WorkOS org membership as the access boundary for the new connections surface.\nMemory\nDecision: Use WorkOS org membership as the access boundary. Scope: project. Source: Linear issue. Future agents should preserve this unless an admin access decision supersedes it.\nLinear connection\nCompany context from Linear\nHow scoped memory works for teams"},{"type":"example","title":"Example: GitHub PR to engineering memory","description":"How pull request reasoning becomes source-linked engineering context for agents.","path":"/examples/github-pr-to-engineering-memory","url":"https://ambience.sh/examples/github-pr-to-engineering-memory","markdownPath":"/examples/github-pr-to-engineering-memory.md","updated":"2026-06-07","summary":"A merged PR tells you what changed. An Ambience memory tells future agents why it changed and what rule to preserve.","proofPoints":["PR descriptions and reviews often contain durable reasoning.","Ambience captures the convention, failure, or decision.","Future agents can retrieve it before touching the same code path."],"sections":[{"title":"Source","body":["GitHub PR: middleware public routes for new marketing pages were covered by route tests to prevent apex redirects from breaking discovery."]},{"title":"Memory","body":["Convention: Public marketing route prefixes must be added to middleware and route tests together. Scope: project. Source: GitHub PR."]}],"related":[{"label":"GitHub connection","href":"/connections/github"},{"label":"Company context for Claude Code","href":"/company-context/for-claude-code"},{"label":"Ambience for Codex","href":"/agents/codex"}],"searchText":"Example: GitHub PR to engineering memory\nHow pull request reasoning becomes source-linked engineering context for agents.\nA merged PR tells you what changed. An Ambience memory tells future agents why it changed and what rule to preserve.\nPR descriptions and reviews often contain durable reasoning.\nAmbience captures the convention, failure, or decision.\nFuture agents can retrieve it before touching the same code path.\nSource\nGitHub PR: middleware public routes for new marketing pages were covered by route tests to prevent apex redirects from breaking discovery.\nMemory\nConvention: Public marketing route prefixes must be added to middleware and route tests together. Scope: project. Source: GitHub PR.\nGitHub connection\nCompany context for Claude Code\nAmbience for Codex"},{"type":"example","title":"Example: Slack thread to team convention","description":"How a Slack decision becomes a reusable Ambience convention for future agents.","path":"/examples/slack-thread-to-team-convention","url":"https://ambience.sh/examples/slack-thread-to-team-convention","markdownPath":"/examples/slack-thread-to-team-convention.md","updated":"2026-06-07","summary":"A Slack thread becomes company context when the durable convention is saved, scoped, and linked back to the source.","proofPoints":["Threads are easy to lose.","Conventions are high-leverage memory.","Ambience stores the reusable rule, not the entire conversation."],"sections":[{"title":"Source","body":["Slack thread: support leads agree that every escalated handoff must include owner, timeframe, and next action."]},{"title":"Memory","body":["Convention: Escalated support handoffs must name owner, timeframe, and next action. Scope: team. Source: Slack thread."]}],"related":[{"label":"Slack connection","href":"/connections/slack"},{"label":"Company context from Slack","href":"/company-context/from-slack"},{"label":"Reusable team skills","href":"/skills"}],"searchText":"Example: Slack thread to team convention\nHow a Slack decision becomes a reusable Ambience convention for future agents.\nA Slack thread becomes company context when the durable convention is saved, scoped, and linked back to the source.\nThreads are easy to lose.\nConventions are high-leverage memory.\nAmbience stores the reusable rule, not the entire conversation.\nSource\nSlack thread: support leads agree that every escalated handoff must include owner, timeframe, and next action.\nMemory\nConvention: Escalated support handoffs must name owner, timeframe, and next action. Scope: team. Source: Slack thread.\nSlack connection\nCompany context from Slack\nReusable team skills"},{"type":"example","title":"Example: new teammate agent onboarding","description":"How Ambience helps a new teammate's agent inherit trusted company context.","path":"/examples/new-teammate-agent-onboarding","url":"https://ambience.sh/examples/new-teammate-agent-onboarding","markdownPath":"/examples/new-teammate-agent-onboarding.md","updated":"2026-06-07","summary":"A new teammate's first agent session receives the team's current conventions, decisions, failures, and references with scopes intact.","proofPoints":["Onboarding context is selected, not dumped.","Sensitive memories remain gated.","The new agent can act with the team's existing operating model."],"sections":[{"title":"First session","body":["The teammate connects Claude Code or Codex. Ambience loads project conventions, active decisions, and known failure modes that the teammate is allowed to see."]},{"title":"After the first week","body":["When the teammate's agent learns something durable, it can save a memory back to the team so the next new session starts smarter too."]}],"related":[{"label":"Onboarding new agents","href":"/company-context/onboarding-new-agents"},{"label":"How to stop agents starting from zero","href":"/answers/how-to-stop-agents-starting-from-zero"},{"label":"Install Ambience","href":"/install"}],"searchText":"Example: new teammate agent onboarding\nHow Ambience helps a new teammate's agent inherit trusted company context.\nA new teammate's first agent session receives the team's current conventions, decisions, failures, and references with scopes intact.\nOnboarding context is selected, not dumped.\nSensitive memories remain gated.\nThe new agent can act with the team's existing operating model.\nFirst session\nThe teammate connects Claude Code or Codex. Ambience loads project conventions, active decisions, and known failure modes that the teammate is allowed to see.\nAfter the first week\nWhen the teammate's agent learns something durable, it can save a memory back to the team so the next new session starts smarter too.\nOnboarding new agents\nHow to stop agents starting from zero\nInstall Ambience"},{"type":"glossary","title":"Company context","description":"Definition of company context for AI agents and why Ambience treats it as governed memory.","path":"/glossary/company-context","url":"https://ambience.sh/glossary/company-context","markdownPath":"/glossary/company-context.md","updated":"2026-06-07","summary":"Company context is the operational memory an AI agent needs to work inside a specific business: decisions, conventions, caveats, sources, and access rules.","proofPoints":["It is created through work, not only documents.","It must be current and source-linked.","It needs permission boundaries before agents reuse it."],"sections":[{"title":"Ambience's view","body":["Ambience treats company context as a working layer, broader than agent memory and more operational than a knowledge base."]}],"related":[{"label":"Company context hub","href":"/company-context"},{"label":"What is company context?","href":"/company-context/what-is-company-context"},{"label":"State of company context report","href":"/research/state-of-company-context-for-ai-agents-2026"}],"searchText":"Company context\nDefinition of company context for AI agents and why Ambience treats it as governed memory.\nCompany context is the operational memory an AI agent needs to work inside a specific business: decisions, conventions, caveats, sources, and access rules.\nIt is created through work, not only documents.\nIt must be current and source-linked.\nIt needs permission boundaries before agents reuse it.\nAmbience's view\nAmbience treats company context as a working layer, broader than agent memory and more operational than a knowledge base.\nCompany context hub\nWhat is company context?\nState of company context report"},{"type":"glossary","title":"Agent memory","description":"Definition of agent memory and how it differs from company context.","path":"/glossary/agent-memory","url":"https://ambience.sh/glossary/agent-memory","markdownPath":"/glossary/agent-memory.md","updated":"2026-06-07","summary":"Agent memory is persistent context an AI agent can retrieve across sessions. Company context adds team governance, source, and access control.","proofPoints":["Personal memory helps one agent recall.","Shared memory needs stronger controls.","Ambience focuses on governed memory for teams."],"sections":[{"title":"Ambience's view","body":["Agent memory becomes company context only when it can be shared safely across people, projects, and approved agents."]}],"related":[{"label":"Company context vs agent memory","href":"/company-context/company-context-vs-agent-memory"},{"label":"Ambience vs Mem0","href":"/vs/mem0"},{"label":"What is governed agent memory?","href":"/answers/what-is-governed-agent-memory"}],"searchText":"Agent memory\nDefinition of agent memory and how it differs from company context.\nAgent memory is persistent context an AI agent can retrieve across sessions. Company context adds team governance, source, and access control.\nPersonal memory helps one agent recall.\nShared memory needs stronger controls.\nAmbience focuses on governed memory for teams.\nAmbience's view\nAgent memory becomes company context only when it can be shared safely across people, projects, and approved agents.\nCompany context vs agent memory\nAmbience vs Mem0\nWhat is governed agent memory?"},{"type":"glossary","title":"Team context","description":"Definition of team context for AI agents.","path":"/glossary/team-context","url":"https://ambience.sh/glossary/team-context","markdownPath":"/glossary/team-context.md","updated":"2026-06-07","summary":"Team context is shared working knowledge for every approved agent working with a team.","proofPoints":["Team context includes conventions, playbooks, decisions, and failures.","It is scoped to the team that owns it.","Ambience makes team context available through governed memory."],"sections":[{"title":"Ambience's view","body":["Team context is the everyday form of company context. It is where review styles, release rules, support practices, and implementation habits become reusable."]}],"related":[{"label":"How scoped memory works for teams","href":"/writing/scoped-memory-for-teams"},{"label":"Reusable team skills","href":"/skills"},{"label":"Company context","href":"/company-context"}],"searchText":"Team context\nDefinition of team context for AI agents.\nTeam context is shared working knowledge for every approved agent working with a team.\nTeam context includes conventions, playbooks, decisions, and failures.\nIt is scoped to the team that owns it.\nAmbience makes team context available through governed memory.\nAmbience's view\nTeam context is the everyday form of company context. It is where review styles, release rules, support practices, and implementation habits become reusable.\nHow scoped memory works for teams\nReusable team skills\nCompany context"},{"type":"glossary","title":"Institutional memory","description":"Definition of institutional memory in the context of AI agents.","path":"/glossary/institutional-memory","url":"https://ambience.sh/glossary/institutional-memory","markdownPath":"/glossary/institutional-memory.md","updated":"2026-06-07","summary":"Institutional memory is what a company has learned over time. Ambience turns the agent-relevant parts into governed, source-linked memory.","proofPoints":["It includes decisions and lessons, not only facts.","It can decay when people leave or tools fragment.","Agents need institutional memory in a retrievable, permissioned form."],"sections":[{"title":"Ambience's view","body":["Institutional memory becomes useful for agents when it is captured as current, scoped, auditable context."]}],"related":[{"label":"How organisational context grows","href":"/writing/growing-organisational-context"},{"label":"How to stop agents starting from zero","href":"/answers/how-to-stop-agents-starting-from-zero"},{"label":"Company context","href":"/company-context"}],"searchText":"Institutional memory\nDefinition of institutional memory in the context of AI agents.\nInstitutional memory is what a company has learned over time. Ambience turns the agent-relevant parts into governed, source-linked memory.\nIt includes decisions and lessons, not only facts.\nIt can decay when people leave or tools fragment.\nAgents need institutional memory in a retrievable, permissioned form.\nAmbience's view\nInstitutional memory becomes useful for agents when it is captured as current, scoped, auditable context.\nHow organisational context grows\nHow to stop agents starting from zero\nCompany context"},{"type":"glossary","title":"Source-linked memory","description":"Definition of source-linked memory for AI agents.","path":"/glossary/source-linked-memory","url":"https://ambience.sh/glossary/source-linked-memory","markdownPath":"/glossary/source-linked-memory.md","updated":"2026-06-07","summary":"Source-linked memory is a durable memory that points back to the call, ticket, PR, document, thread, or session that produced it.","proofPoints":["The source makes the memory inspectable.","The memory stays concise.","The agent can cite where the decision came from."],"sections":[{"title":"Ambience's view","body":["Source-linked memory is how Ambience avoids turning company context into unsourced prompt folklore."]}],"related":[{"label":"Source-linked decision example","href":"/examples/source-linked-decision"},{"label":"Granola connection","href":"/connections/granola"},{"label":"How to audit AI agent context","href":"/answers/how-to-audit-ai-agent-context"}],"searchText":"Source-linked memory\nDefinition of source-linked memory for AI agents.\nSource-linked memory is a durable memory that points back to the call, ticket, PR, document, thread, or session that produced it.\nThe source makes the memory inspectable.\nThe memory stays concise.\nThe agent can cite where the decision came from.\nAmbience's view\nSource-linked memory is how Ambience avoids turning company context into unsourced prompt folklore.\nSource-linked decision example\nGranola connection\nHow to audit AI agent context"},{"type":"glossary","title":"Governed memory","description":"Definition of governed memory for AI agents.","path":"/glossary/governed-memory","url":"https://ambience.sh/glossary/governed-memory","markdownPath":"/glossary/governed-memory.md","updated":"2026-06-07","summary":"Governed memory is persistent agent context with controls: source, scope, redaction, conflict review, and audit.","proofPoints":["It constrains retrieval before relevance ranking is enough.","It preserves who can read what.","Ambience is built around governed memory for companies."],"sections":[{"title":"Ambience's view","body":["Governed memory is what makes company context safe enough for work. It lets agents reuse context without flattening permissions."]}],"related":[{"label":"What is governed agent memory?","href":"/answers/what-is-governed-agent-memory"},{"label":"Security and permissions","href":"/company-context/security-and-permissions"},{"label":"Security and trust","href":"/security"}],"searchText":"Governed memory\nDefinition of governed memory for AI agents.\nGoverned memory is persistent agent context with controls: source, scope, redaction, conflict review, and audit.\nIt constrains retrieval before relevance ranking is enough.\nIt preserves who can read what.\nAmbience is built around governed memory for companies.\nAmbience's view\nGoverned memory is what makes company context safe enough for work. It lets agents reuse context without flattening permissions.\nWhat is governed agent memory?\nSecurity and permissions\nSecurity and trust"},{"type":"glossary","title":"Scoped memory","description":"Definition of scoped memory for teams using AI agents.","path":"/glossary/scoped-memory","url":"https://ambience.sh/glossary/scoped-memory","markdownPath":"/glossary/scoped-memory.md","updated":"2026-06-07","summary":"Scoped memory is agent context that belongs to a specific access boundary: personal, team, project, org, or sensitive.","proofPoints":["Scopes prevent private context from becoming shared by accident.","Project memories follow project work.","Sensitive memories require explicit grants."],"sections":[{"title":"Ambience's view","body":["Scoped memory is the basic unit of company context governance in Ambience."]}],"related":[{"label":"How scoped memory works for teams","href":"/writing/scoped-memory-for-teams"},{"label":"Company context from Linear","href":"/company-context/from-linear"},{"label":"Security and permissions","href":"/company-context/security-and-permissions"}],"searchText":"Scoped memory\nDefinition of scoped memory for teams using AI agents.\nScoped memory is agent context that belongs to a specific access boundary: personal, team, project, org, or sensitive.\nScopes prevent private context from becoming shared by accident.\nProject memories follow project work.\nSensitive memories require explicit grants.\nAmbience's view\nScoped memory is the basic unit of company context governance in Ambience.\nHow scoped memory works for teams\nCompany context from Linear\nSecurity and permissions"},{"type":"glossary","title":"Context harness","description":"Definition of context harness for AI agents.","path":"/glossary/context-harness","url":"https://ambience.sh/glossary/context-harness","markdownPath":"/glossary/context-harness.md","updated":"2026-06-07","summary":"A context harness is the system that captures, filters, and supplies the right context to an AI agent at the right moment.","proofPoints":["It is more than a prompt.","It includes capture, retrieval, permissions, source, and feedback.","Ambience is a company context harness for approved agents."],"sections":[{"title":"Ambience's view","body":["A context harness makes every agent session smarter without making every source globally visible."]}],"related":[{"label":"Company context for AI agents","href":"/company-context/for-ai-agents"},{"label":"State of company context report","href":"/research/state-of-company-context-for-ai-agents-2026"},{"label":"Install Ambience","href":"/install"}],"searchText":"Context harness\nDefinition of context harness for AI agents.\nA context harness is the system that captures, filters, and supplies the right context to an AI agent at the right moment.\nIt is more than a prompt.\nIt includes capture, retrieval, permissions, source, and feedback.\nAmbience is a company context harness for approved agents.\nAmbience's view\nA context harness makes every agent session smarter without making every source globally visible.\nCompany context for AI agents\nState of company context report\nInstall Ambience"},{"type":"glossary","title":"MCP memory","description":"Definition of MCP memory for AI agents.","path":"/glossary/mcp-memory","url":"https://ambience.sh/glossary/mcp-memory","markdownPath":"/glossary/mcp-memory.md","updated":"2026-06-07","summary":"MCP memory is memory exposed to AI agents through Model Context Protocol tools, so agents can search, read, and save context with explicit tool calls.","proofPoints":["MCP makes memory agent-native.","Tool access needs the same scope and audit controls as the app.","Ambience exposes memory workflows through MCP and CLI surfaces."],"sections":[{"title":"Ambience's view","body":["MCP is the transport. Ambience is the governed memory plane behind it."]}],"related":[{"label":"Ambience for Cursor","href":"/agents/cursor"},{"label":"Ambience for Codex","href":"/agents/codex"},{"label":"What is governed agent memory?","href":"/answers/what-is-governed-agent-memory"}],"searchText":"MCP memory\nDefinition of MCP memory for AI agents.\nMCP memory is memory exposed to AI agents through Model Context Protocol tools, so agents can search, read, and save context with explicit tool calls.\nMCP makes memory agent-native.\nTool access needs the same scope and audit controls as the app.\nAmbience exposes memory workflows through MCP and CLI surfaces.\nAmbience's view\nMCP is the transport. Ambience is the governed memory plane behind it.\nAmbience for Cursor\nAmbience for Codex\nWhat is governed agent memory?"},{"type":"glossary","title":"Agent onboarding","description":"Definition of agent onboarding with company context.","path":"/glossary/agent-onboarding","url":"https://ambience.sh/glossary/agent-onboarding","markdownPath":"/glossary/agent-onboarding.md","updated":"2026-06-07","summary":"Agent onboarding is the process of giving a newly approved agent the company context it needs without leaking private or stale information.","proofPoints":["The user chooses included sources.","The system seeds reviewed memories.","Future sessions show context reuse."],"sections":[{"title":"Ambience's view","body":["Good agent onboarding is one visible connection, one confirmed first memory, and enough context to make the next session useful."]}],"related":[{"label":"Onboarding new agents","href":"/company-context/onboarding-new-agents"},{"label":"New teammate agent onboarding","href":"/examples/new-teammate-agent-onboarding"},{"label":"Install Ambience","href":"/install"}],"searchText":"Agent onboarding\nDefinition of agent onboarding with company context.\nAgent onboarding is the process of giving a newly approved agent the company context it needs without leaking private or stale information.\nThe user chooses included sources.\nThe system seeds reviewed memories.\nFuture sessions show context reuse.\nAmbience's view\nGood agent onboarding is one visible connection, one confirmed first memory, and enough context to make the next session useful.\nOnboarding new agents\nNew teammate agent onboarding\nInstall Ambience"},{"type":"glossary","title":"Memory conflict review","description":"Definition of memory conflict review for company context.","path":"/glossary/memory-conflict-review","url":"https://ambience.sh/glossary/memory-conflict-review","markdownPath":"/glossary/memory-conflict-review.md","updated":"2026-06-07","summary":"Memory conflict review is the process of resolving stale, contradictory, or scope-mismatched memories before agents reuse them.","proofPoints":["Agent memory can propagate errors if old experience is replayed blindly.","Conflict review lets humans decide which context is current.","Ambience records those decisions so retrieval improves over time."],"sections":[{"title":"Ambience's view","body":["Company context learns from disagreement. Keeping stale memory invisible is not enough; the resolution itself is durable context."]}],"related":[{"label":"How organisational context grows","href":"/writing/growing-organisational-context"},{"label":"State of company context report","href":"/research/state-of-company-context-for-ai-agents-2026"},{"label":"Security and permissions","href":"/company-context/security-and-permissions"}],"searchText":"Memory conflict review\nDefinition of memory conflict review for company context.\nMemory conflict review is the process of resolving stale, contradictory, or scope-mismatched memories before agents reuse them.\nAgent memory can propagate errors if old experience is replayed blindly.\nConflict review lets humans decide which context is current.\nAmbience records those decisions so retrieval improves over time.\nAmbience's view\nCompany context learns from disagreement. Keeping stale memory invisible is not enough; the resolution itself is durable context.\nHow organisational context grows\nState of company context report\nSecurity and permissions"},{"type":"comparison","title":"Ambience vs Glean for company context","description":"Glean is enterprise search. Ambience is governed company context for AI agents.","path":"/compare/glean","url":"https://ambience.sh/compare/glean","markdownPath":"/compare/glean.md","updated":"2026-06-07","summary":"Ambience is the stronger fit when agents need writable, source-linked, scoped memory from work, not only enterprise search over existing knowledge.","proofPoints":["Search finds documents; Ambience captures durable agent-facing memory.","Ambience can write new context as work happens.","Scopes, redaction, and conflict review are first-class in Ambience."],"sections":[{"title":"The category difference","body":["Enterprise search answers 'where is the information?' Company context answers 'what does this agent need before acting, and is it allowed to know it?'"]}],"related":[{"label":"Company context vs knowledge base","href":"/company-context/company-context-vs-knowledge-base"},{"label":"What is company context?","href":"/company-context/what-is-company-context"},{"label":"Security and trust","href":"/security"}],"searchText":"Ambience vs Glean for company context\nGlean is enterprise search. Ambience is governed company context for AI agents.\nAmbience is the stronger fit when agents need writable, source-linked, scoped memory from work, not only enterprise search over existing knowledge.\nSearch finds documents; Ambience captures durable agent-facing memory.\nAmbience can write new context as work happens.\nScopes, redaction, and conflict review are first-class in Ambience.\nThe category difference\nEnterprise search answers 'where is the information?' Company context answers 'what does this agent need before acting, and is it allowed to know it?'\nCompany context vs knowledge base\nWhat is company context?\nSecurity and trust"},{"type":"comparison","title":"Ambience vs Notion AI for company context","description":"Notion AI helps humans use a workspace. Ambience gives agents governed company context.","path":"/compare/notion-ai","url":"https://ambience.sh/compare/notion-ai","markdownPath":"/compare/notion-ai.md","updated":"2026-06-07","summary":"Ambience is stronger when the context needs to move from Notion and other tools into approved agents with source, scope, and audit.","proofPoints":["Ambience can use Notion as a source without making Notion the agent memory plane.","Skills and decisions become callable context across agent runtimes.","Governance stays attached outside the original workspace."],"sections":[{"title":"The workflow difference","body":["Notion AI improves a Notion page. Ambience extracts the durable context agents need and makes it reusable across Claude Code, Codex, Cursor, Copilot, and MCP tools."]}],"related":[{"label":"Notion connection","href":"/connections/notion"},{"label":"Company context vs knowledge base","href":"/company-context/company-context-vs-knowledge-base"},{"label":"Reusable team skills","href":"/skills"}],"searchText":"Ambience vs Notion AI for company context\nNotion AI helps humans use a workspace. Ambience gives agents governed company context.\nAmbience is stronger when the context needs to move from Notion and other tools into approved agents with source, scope, and audit.\nAmbience can use Notion as a source without making Notion the agent memory plane.\nSkills and decisions become callable context across agent runtimes.\nGovernance stays attached outside the original workspace.\nThe workflow difference\nNotion AI improves a Notion page. Ambience extracts the durable context agents need and makes it reusable across Claude Code, Codex, Cursor, Copilot, and MCP tools.\nNotion connection\nCompany context vs knowledge base\nReusable team skills"},{"type":"comparison","title":"Ambience vs Confluence for company context","description":"Confluence stores human documentation. Ambience stores governed memory for AI agents.","path":"/compare/confluence","url":"https://ambience.sh/compare/confluence","markdownPath":"/compare/confluence.md","updated":"2026-06-07","summary":"Ambience is stronger when agents need current decisions, conventions, and failures with source and access controls, not another page archive.","proofPoints":["Documentation is often stale by the time an agent needs to act.","Ambience captures live work context as typed memories.","Source links preserve the original page or ticket for inspection."],"sections":[{"title":"Docs are still useful","body":["Ambience does not replace documentation. It turns the agent-relevant pieces into governed context that can be loaded at session start."]}],"related":[{"label":"Company context vs knowledge base","href":"/company-context/company-context-vs-knowledge-base"},{"label":"Google Docs connection","href":"/connections/google-docs"},{"label":"What is governed agent memory?","href":"/answers/what-is-governed-agent-memory"}],"searchText":"Ambience vs Confluence for company context\nConfluence stores human documentation. Ambience stores governed memory for AI agents.\nAmbience is stronger when agents need current decisions, conventions, and failures with source and access controls, not another page archive.\nDocumentation is often stale by the time an agent needs to act.\nAmbience captures live work context as typed memories.\nSource links preserve the original page or ticket for inspection.\nDocs are still useful\nAmbience does not replace documentation. It turns the agent-relevant pieces into governed context that can be loaded at session start.\nCompany context vs knowledge base\nGoogle Docs connection\nWhat is governed agent memory?"},{"type":"comparison","title":"Ambience vs Guru for company context","description":"Guru helps teams manage knowledge cards. Ambience gives AI agents governed, source-linked company context.","path":"/compare/guru","url":"https://ambience.sh/compare/guru","markdownPath":"/compare/guru.md","updated":"2026-06-07","summary":"Ambience is stronger when knowledge must be captured from agent work and reused by future agents under scope and audit controls.","proofPoints":["Ambience memories are written for agent retrieval.","Agent sessions can create new durable context.","Conflict review handles changing decisions."],"sections":[{"title":"The verification problem","body":["Human-facing verification is useful. Agent-facing context also needs source, permissions, redaction, and a retrieval boundary that applies every time."]}],"related":[{"label":"Company context for AI agents","href":"/company-context/for-ai-agents"},{"label":"Memory conflict review","href":"/glossary/memory-conflict-review"},{"label":"Security and trust","href":"/security"}],"searchText":"Ambience vs Guru for company context\nGuru helps teams manage knowledge cards. Ambience gives AI agents governed, source-linked company context.\nAmbience is stronger when knowledge must be captured from agent work and reused by future agents under scope and audit controls.\nAmbience memories are written for agent retrieval.\nAgent sessions can create new durable context.\nConflict review handles changing decisions.\nThe verification problem\nHuman-facing verification is useful. Agent-facing context also needs source, permissions, redaction, and a retrieval boundary that applies every time.\nCompany context for AI agents\nMemory conflict review\nSecurity and trust"},{"type":"comparison","title":"Ambience vs Slack AI for company context","description":"Slack AI can summarize conversations. Ambience turns important decisions into governed memory for all approved agents.","path":"/compare/slack-ai","url":"https://ambience.sh/compare/slack-ai","markdownPath":"/compare/slack-ai.md","updated":"2026-06-07","summary":"Ambience is stronger when Slack decisions need to survive as scoped context across agent runtimes.","proofPoints":["The Slack thread remains the source.","Ambience stores the reusable decision or convention.","Future agents can retrieve it outside Slack with scope and audit."],"sections":[{"title":"The thread is not the memory","body":["The durable artifact is usually a decision, owner, risk, or convention. Ambience stores that concise memory and links back to Slack as provenance."]}],"related":[{"label":"Slack connection","href":"/connections/slack"},{"label":"Company context from Slack","href":"/company-context/from-slack"},{"label":"Slack thread example","href":"/examples/slack-thread-to-team-convention"}],"searchText":"Ambience vs Slack AI for company context\nSlack AI can summarize conversations. Ambience turns important decisions into governed memory for all approved agents.\nAmbience is stronger when Slack decisions need to survive as scoped context across agent runtimes.\nThe Slack thread remains the source.\nAmbience stores the reusable decision or convention.\nFuture agents can retrieve it outside Slack with scope and audit.\nThe thread is not the memory\nThe durable artifact is usually a decision, owner, risk, or convention. Ambience stores that concise memory and links back to Slack as provenance.\nSlack connection\nCompany context from Slack\nSlack thread example"},{"type":"comparison","title":"Ambience vs Google Drive and Gemini for company context","description":"Google Drive and Gemini help users search and summarize workspace files. Ambience gives agents governed context across tools.","path":"/compare/google-drive-gemini","url":"https://ambience.sh/compare/google-drive-gemini","markdownPath":"/compare/google-drive-gemini.md","updated":"2026-06-07","summary":"Ambience is stronger when the output of documents needs to become reusable, source-linked memory for agents outside a single workspace surface.","proofPoints":["Ambience can use docs as sources.","The durable memory is shorter than the source document.","Access and audit remain attached to future agent retrieval."],"sections":[{"title":"From document to memory","body":["An RFC can stay in Drive. Ambience captures the decisions and constraints that should shape future agent work."]}],"related":[{"label":"Google Docs connection","href":"/connections/google-docs"},{"label":"Company context vs knowledge base","href":"/company-context/company-context-vs-knowledge-base"},{"label":"Security and trust","href":"/security"}],"searchText":"Ambience vs Google Drive and Gemini for company context\nGoogle Drive and Gemini help users search and summarize workspace files. Ambience gives agents governed context across tools.\nAmbience is stronger when the output of documents needs to become reusable, source-linked memory for agents outside a single workspace surface.\nAmbience can use docs as sources.\nThe durable memory is shorter than the source document.\nAccess and audit remain attached to future agent retrieval.\nFrom document to memory\nAn RFC can stay in Drive. Ambience captures the decisions and constraints that should shape future agent work.\nGoogle Docs connection\nCompany context vs knowledge base\nSecurity and trust"},{"type":"comparison","title":"Ambience vs LangSmith and LangGraph memory","description":"LangSmith and LangGraph help teams build agents. Ambience gives working teams a governed memory plane across agents.","path":"/compare/langsmith-langgraph-memory","url":"https://ambience.sh/compare/langsmith-langgraph-memory","markdownPath":"/compare/langsmith-langgraph-memory.md","updated":"2026-06-07","summary":"Ambience is stronger when the problem is shared company context for humans and agents, not only application-agent memory inside a framework.","proofPoints":["Framework memory is usually application-specific.","Ambience is runtime-agnostic across developer agents and MCP clients.","Company context needs UI, CLI, MCP, onboarding, permissions, and audit together."],"sections":[{"title":"Framework memory vs company context","body":["A framework can help one agent app remember. Ambience helps an organisation share context across the agents people already use."]}],"related":[{"label":"MCP memory","href":"/glossary/mcp-memory"},{"label":"Ambience for Cursor","href":"/agents/cursor"},{"label":"Company context for AI agents","href":"/company-context/for-ai-agents"}],"searchText":"Ambience vs LangSmith and LangGraph memory\nLangSmith and LangGraph help teams build agents. Ambience gives working teams a governed memory plane across agents.\nAmbience is stronger when the problem is shared company context for humans and agents, not only application-agent memory inside a framework.\nFramework memory is usually application-specific.\nAmbience is runtime-agnostic across developer agents and MCP clients.\nCompany context needs UI, CLI, MCP, onboarding, permissions, and audit together.\nFramework memory vs company context\nA framework can help one agent app remember. Ambience helps an organisation share context across the agents people already use.\nMCP memory\nAmbience for Cursor\nCompany context for AI agents"},{"type":"comparison","title":"Ambience vs internal RAG for company context","description":"Internal RAG retrieves documents. Ambience captures and governs live company context for AI agents.","path":"/compare/internal-rag","url":"https://ambience.sh/compare/internal-rag","markdownPath":"/compare/internal-rag.md","updated":"2026-06-07","summary":"Ambience is stronger when the context is created by ongoing work and must be scoped, redacted, audited, and updated over time.","proofPoints":["RAG can retrieve stale documents.","Ambience can capture new decisions as they happen.","Conflict review helps resolve changing context."],"sections":[{"title":"Retrieval is necessary but not sufficient","body":["Company context is not just a bigger corpus. It needs write paths, source evidence, permissions, and a way to correct stale memory."]}],"related":[{"label":"Memory conflict review","href":"/glossary/memory-conflict-review"},{"label":"How organisational context grows","href":"/writing/growing-organisational-context"},{"label":"State of company context report","href":"/research/state-of-company-context-for-ai-agents-2026"}],"searchText":"Ambience vs internal RAG for company context\nInternal RAG retrieves documents. Ambience captures and governs live company context for AI agents.\nAmbience is stronger when the context is created by ongoing work and must be scoped, redacted, audited, and updated over time.\nRAG can retrieve stale documents.\nAmbience can capture new decisions as they happen.\nConflict review helps resolve changing context.\nRetrieval is necessary but not sufficient\nCompany context is not just a bigger corpus. It needs write paths, source evidence, permissions, and a way to correct stale memory.\nMemory conflict review\nHow organisational context grows\nState of company context report"},{"type":"comparison","title":"Ambience vs vector databases for company context","description":"Vector databases store embeddings. Ambience provides the governed memory workflow around company context.","path":"/compare/vector-databases","url":"https://ambience.sh/compare/vector-databases","markdownPath":"/compare/vector-databases.md","updated":"2026-06-07","summary":"Ambience is stronger when teams need a product layer for capture, scope, source, redaction, audit, and agent access, not only a retrieval primitive.","proofPoints":["Embeddings do not decide who can read a memory.","A database does not create source-linked decisions from work.","Ambience can sit above retrieval primitives as the governed memory plane."],"sections":[{"title":"The primitive is not the product","body":["A vector database can be part of a memory system. Ambience is the user-facing, agent-facing context system teams need around storage."]}],"related":[{"label":"What is governed agent memory?","href":"/answers/what-is-governed-agent-memory"},{"label":"Company context vs agent memory","href":"/company-context/company-context-vs-agent-memory"},{"label":"Security and permissions","href":"/company-context/security-and-permissions"}],"searchText":"Ambience vs vector databases for company context\nVector databases store embeddings. Ambience provides the governed memory workflow around company context.\nAmbience is stronger when teams need a product layer for capture, scope, source, redaction, audit, and agent access, not only a retrieval primitive.\nEmbeddings do not decide who can read a memory.\nA database does not create source-linked decisions from work.\nAmbience can sit above retrieval primitives as the governed memory plane.\nThe primitive is not the product\nA vector database can be part of a memory system. Ambience is the user-facing, agent-facing context system teams need around storage.\nWhat is governed agent memory?\nCompany context vs agent memory\nSecurity and permissions"},{"type":"comparison","title":"Ambience vs Mem0","description":"Mem0 is built for single-agent recall. Ambience is built for governed team context: scoped, redacted, audited memory for agents at work.","path":"/vs/mem0","url":"https://ambience.sh/vs/mem0","markdownPath":"/vs/mem0.md","updated":"2026-06-07","summary":"Ambience is the better fit when memory must be shared across a team with policy-enforced scopes, redaction, audit, and revocation.","sections":[{"title":"What Mem0 optimizes for","body":["Mem0 is strongest for single-agent recall."]},{"title":"Why Ambience is stronger for team context","body":["Ambience adds team scopes, redaction before storage, SSO, revocation, and audit."]},{"title":"Side-by-side","body":["Capability comparison for team memory governance."]}],"related":[{"label":"Single-agent memory vs team context","href":"/writing/agent-memory-vs-team-context"}],"searchText":"Ambience vs Mem0\nMem0 is built for single-agent recall. Ambience is built for governed team context: scoped, redacted, audited memory for agents at work.\nAmbience is the better fit when memory must be shared across a team with policy-enforced scopes, redaction, audit, and revocation.\nWhat Mem0 optimizes for\nMem0 is strongest for single-agent recall.\nWhy Ambience is stronger for team context\nAmbience adds team scopes, redaction before storage, SSO, revocation, and audit.\nSide-by-side\nCapability comparison for team memory governance.\nSingle-agent memory vs team context"},{"type":"comparison","title":"Ambience vs Zep","description":"Zep is a context-engineering platform built on a temporal knowledge graph. Ambience is built for governed team memory with scope, redaction, identity, and audit.","path":"/vs/zep","url":"https://ambience.sh/vs/zep","markdownPath":"/vs/zep.md","updated":"2026-06-07","summary":"Ambience is the better fit when shared agent context must be governed by people, projects, and permissions.","sections":[{"title":"What Zep optimizes for","body":["Zep focuses on temporal graph memory for application agents."]},{"title":"Why Ambience is stronger for team context","body":["Ambience focuses on company context across working agents and teammates."]}],"related":[{"label":"Company context vs agent memory","href":"/company-context/company-context-vs-agent-memory"}],"searchText":"Ambience vs Zep\nZep is a context-engineering platform built on a temporal knowledge graph. Ambience is built for governed team memory with scope, redaction, identity, and audit.\nAmbience is the better fit when shared agent context must be governed by people, projects, and permissions.\nWhat Zep optimizes for\nZep focuses on temporal graph memory for application agents.\nWhy Ambience is stronger for team context\nAmbience focuses on company context across working agents and teammates.\nCompany context vs agent memory"},{"type":"comparison","title":"Ambience vs Letta","description":"Letta is built for stateful agent architecture. Ambience is built for organisation-wide context governance.","path":"/vs/letta","url":"https://ambience.sh/vs/letta","markdownPath":"/vs/letta.md","updated":"2026-06-07","summary":"Ambience is the better fit when teams need a governed memory plane outside one agent architecture.","sections":[{"title":"What Letta optimizes for","body":["Letta focuses on long-running stateful agents."]},{"title":"Why Ambience is stronger for team context","body":["Ambience focuses on shared context controls, source links, redaction, and audit."]}],"related":[{"label":"Company context for AI agents","href":"/company-context/for-ai-agents"}],"searchText":"Ambience vs Letta\nLetta is built for stateful agent architecture. Ambience is built for organisation-wide context governance.\nAmbience is the better fit when teams need a governed memory plane outside one agent architecture.\nWhat Letta optimizes for\nLetta focuses on long-running stateful agents.\nWhy Ambience is stronger for team context\nAmbience focuses on shared context controls, source links, redaction, and audit.\nCompany context for AI agents"},{"type":"comparison","title":"Ambience vs Hyper","description":"Hyper is a self-driving company brain. Ambience is the governed memory plane for reusable context with scope, redaction, conflict review, audit, and proof.","path":"/vs/hyper","url":"https://ambience.sh/vs/hyper","markdownPath":"/vs/hyper.md","updated":"2026-06-07","summary":"Ambience is the better fit when teams want focused memory governance rather than an invisible broad company brain.","sections":[{"title":"What Hyper optimizes for","body":["Hyper positions as an invisible company brain."]},{"title":"Why Ambience is stronger for team context","body":["Ambience makes context capture, source, scope, and proof visible."]}],"related":[{"label":"State of company context report","href":"/research/state-of-company-context-for-ai-agents-2026"}],"searchText":"Ambience vs Hyper\nHyper is a self-driving company brain. Ambience is the governed memory plane for reusable context with scope, redaction, conflict review, audit, and proof.\nAmbience is the better fit when teams want focused memory governance rather than an invisible broad company brain.\nWhat Hyper optimizes for\nHyper positions as an invisible company brain.\nWhy Ambience is stronger for team context\nAmbience makes context capture, source, scope, and proof visible.\nState of company context report"},{"type":"comparison","title":"Ambience vs Wato","description":"Wato is a broad shared AI workspace. Ambience is narrower by design: a governed memory plane for durable agent context.","path":"/vs/wato","url":"https://ambience.sh/vs/wato","markdownPath":"/vs/wato.md","updated":"2026-06-07","summary":"Ambience is the better fit when the problem is governed memory, not a full AI workspace with tools, workflows, cloud sessions, and traces.","sections":[{"title":"What Wato optimizes for","body":["Wato combines memory, MCP, tools, workflows, cloud sessions, artifacts, and traces."]},{"title":"Why Ambience is stronger for team context","body":["Ambience focuses on source-linked memory, permissions, redaction, and audit."]}],"related":[{"label":"Context harness","href":"/glossary/context-harness"}],"searchText":"Ambience vs Wato\nWato is a broad shared AI workspace. Ambience is narrower by design: a governed memory plane for durable agent context.\nAmbience is the better fit when the problem is governed memory, not a full AI workspace with tools, workflows, cloud sessions, and traces.\nWhat Wato optimizes for\nWato combines memory, MCP, tools, workflows, cloud sessions, artifacts, and traces.\nWhy Ambience is stronger for team context\nAmbience focuses on source-linked memory, permissions, redaction, and audit.\nContext harness"},{"type":"research","title":"State of company context for AI agents, 2026","description":"A research-backed argument for company context as the missing layer between enterprise data and reliable AI agents.","path":"/research/state-of-company-context-for-ai-agents-2026","url":"https://ambience.sh/research/state-of-company-context-for-ai-agents-2026","markdownPath":"/research/state-of-company-context-for-ai-agents-2026.md","updated":"2026-06-07","summary":"The next bottleneck for workplace AI agents is not model intelligence alone. It is company context: the governed, source-linked memory that lets agents act with the organisation's current decisions, conventions, and constraints.","proofPoints":["Search engines and AI answer systems reward crawlable, textual, internally linked, sourceable content.","Research on agent memory shows unmanaged experience replay can propagate errors and stale behavior.","The enterprise answer is governed company context: memory with source, scope, redaction, review, and audit."],"sections":[{"title":"Thesis","body":["Agents are moving from demos into work, and the limiting factor is increasingly the context they receive. A powerful model without company context still starts cold.","The market has treated this as a retrieval problem. That is incomplete. Company context is a governance problem: what should be remembered, where it came from, who can use it, whether it is still current, and how the organisation proves it shaped the agent's action."]},{"title":"Why search visibility now depends on answerability","body":["OpenAI's publisher guidance says public sites can appear in ChatGPT search and should not block OAI-SearchBot if they want summaries and snippets. Google says AI features still depend on ordinary indexing fundamentals: crawlability, internal links, textual content, page experience, and structured data that matches visible content. Bing now exposes AI Performance reporting for cited pages and grounding queries.","That means category leadership needs crisp public answers, not only landing-page claims. Ambience publishes the canonical definition of company context and the operational examples agents can cite."]},{"title":"Why memory alone is not enough","body":["Recent agent-memory research is useful but also a warning. The experience-following study found that agents can repeat patterns from retrieved memory and that low-quality memory can propagate errors or misleading behavior. Episodic-memory research argues that long-term agents need richer memory to learn from specific contexts. Agentic Context Engineering shows that evolving playbooks can improve agent performance when they are curated and updated.","The lesson for companies is blunt: memory needs lifecycle and governance. A pile of recalled experiences is not safe company context."]},{"title":"The company context stack","body":["A serious company context layer has six jobs: capture durable work context, preserve source evidence, redact sensitive text before storage, scope retrieval, resolve stale or conflicting memories, and audit reads and writes.","Ambience is built around that stack. It captures decisions, patterns, skills, conventions, failures, and references from agent sessions and approved sources, then makes them available to future agents under explicit access rules."]},{"title":"The wedge","body":["The fastest path is not building every connector first. Most teams already have agents connected to Slack, Linear, GitHub, Granola, Google Docs, Notion, and local files. Ambience can use those approved agent connections to propose source-linked memories, ask what the user wants included, and seed governed company context without taking blanket credentials up front."]},{"title":"Prediction","body":["In 2026, the winning teams will stop asking whether an agent has memory and start asking whether the company's context is governed. The durable category will not be 'agent memory' by itself. It will be company context for agents."]}],"related":[{"label":"Company context","href":"/company-context"},{"label":"Memory conflict review","href":"/glossary/memory-conflict-review"},{"label":"Security and permissions","href":"/company-context/security-and-permissions"}],"searchText":"State of company context for AI agents, 2026\nA research-backed argument for company context as the missing layer between enterprise data and reliable AI agents.\nThe next bottleneck for workplace AI agents is not model intelligence alone. It is company context: the governed, source-linked memory that lets agents act with the organisation's current decisions, conventions, and constraints.\nSearch engines and AI answer systems reward crawlable, textual, internally linked, sourceable content.\nResearch on agent memory shows unmanaged experience replay can propagate errors and stale behavior.\nThe enterprise answer is governed company context: memory with source, scope, redaction, review, and audit.\nThesis\nAgents are moving from demos into work, and the limiting factor is increasingly the context they receive. A powerful model without company context still starts cold.\nThe market has treated this as a retrieval problem. That is incomplete. Company context is a governance problem: what should be remembered, where it came from, who can use it, whether it is still current, and how the organisation proves it shaped the agent's action.\nWhy search visibility now depends on answerability\nOpenAI's publisher guidance says public sites can appear in ChatGPT search and should not block OAI-SearchBot if they want summaries and snippets. Google says AI features still depend on ordinary indexing fundamentals: crawlability, internal links, textual content, page experience, and structured data that matches visible content. Bing now exposes AI Performance reporting for cited pages and grounding queries.\nThat means category leadership needs crisp public answers, not only landing-page claims. Ambience publishes the canonical definition of company context and the operational examples agents can cite.\nWhy memory alone is not enough\nRecent agent-memory research is useful but also a warning. The experience-following study found that agents can repeat patterns from retrieved memory and that low-quality memory can propagate errors or misleading behavior. Episodic-memory research argues that long-term agents need richer memory to learn from specific contexts. Agentic Context Engineering shows that evolving playbooks can improve agent performance when they are curated and updated.\nThe lesson for companies is blunt: memory needs lifecycle and governance. A pile of recalled experiences is not safe company context.\nThe company context stack\nA serious company context layer has six jobs: capture durable work context, preserve source evidence, redact sensitive text before storage, scope retrieval, resolve stale or conflicting memories, and audit reads and writes.\nAmbience is built around that stack. It captures decisions, patterns, skills, conventions, failures, and references from agent sessions and approved sources, then makes them available to future agents under explicit access rules.\nThe wedge\nThe fastest path is not building every connector first. Most teams already have agents connected to Slack, Linear, GitHub, Granola, Google Docs, Notion, and local files. Ambience can use those approved agent connections to propose source-linked memories, ask what the user wants included, and seed governed company context without taking blanket credentials up front.\nPrediction\nIn 2026, the winning teams will stop asking whether an agent has memory and start asking whether the company's context is governed. The durable category will not be 'agent memory' by itself. It will be company context for agents.\nCompany context\nMemory conflict review\nSecurity and permissions"},{"type":"technical-note","title":"Single-agent memory vs team context","description":"Agent memory covers two products: single-agent recall and governed team context. Ambience is built for the team context layer.","path":"/writing/agent-memory-vs-team-context","url":"https://ambience.sh/writing/agent-memory-vs-team-context","markdownPath":"/writing/agent-memory-vs-team-context.md","updated":"2026-06-07","summary":"Single-agent memory is a retrieval problem. Team context is a governance problem involving scope, redaction, audit, and shared reuse.","sections":[{"title":"What teams actually arrive with","body":["Teams need shared agent context without copy-pasting, leaking secrets, or losing access control."]},{"title":"So how do you choose","body":["Single-agent memory fits private recall. Ambience fits teams that need governed context across people and agents."]},{"title":"Why we built Ambience","body":["Ambience captures sessions, redacts before storage, scopes memory, and audits every read."]}],"related":[{"label":"How organisational context grows","href":"/writing/growing-organisational-context"},{"label":"How scoped memory works for teams","href":"/writing/scoped-memory-for-teams"},{"label":"Ambience vs Mem0","href":"/vs/mem0"}],"searchText":"Single-agent memory vs team context\nAgent memory covers two products: single-agent recall and governed team context. Ambience is built for the team context layer.\nSingle-agent memory is a retrieval problem. Team context is a governance problem involving scope, redaction, audit, and shared reuse.\nWhat teams actually arrive with\nTeams need shared agent context without copy-pasting, leaking secrets, or losing access control.\nSo how do you choose\nSingle-agent memory fits private recall. Ambience fits teams that need governed context across people and agents.\nWhy we built Ambience\nAmbience captures sessions, redacts before storage, scopes memory, and audits every read.\nHow organisational context grows\nHow scoped memory works for teams\nAmbience vs Mem0"},{"type":"technical-note","title":"How organisational context grows","description":"A technical note on memory, conflict resolution, and human feedback loops for AI agents at work.","path":"/writing/growing-organisational-context","url":"https://ambience.sh/writing/growing-organisational-context","markdownPath":"/writing/growing-organisational-context.md","updated":"2026-06-07","summary":"Useful organisational context grows through capture, constraints, claim extraction, conflict review, and human feedback.","sections":[{"title":"Context has a lifecycle","body":["Agent memory needs capture, constraint, review, and reuse."]},{"title":"Claims make memories comparable","body":["Derived claims let Ambience compare context without losing the source memory."]},{"title":"Conflicts are the learning moment","body":["Human conflict review teaches the organisation model what to trust."]}],"related":[{"label":"Memory conflict review","href":"/glossary/memory-conflict-review"},{"label":"How scoped memory works for teams","href":"/writing/scoped-memory-for-teams"},{"label":"Company context report","href":"/research/state-of-company-context-for-ai-agents-2026"}],"searchText":"How organisational context grows\nA technical note on memory, conflict resolution, and human feedback loops for AI agents at work.\nUseful organisational context grows through capture, constraints, claim extraction, conflict review, and human feedback.\nContext has a lifecycle\nAgent memory needs capture, constraint, review, and reuse.\nClaims make memories comparable\nDerived claims let Ambience compare context without losing the source memory.\nConflicts are the learning moment\nHuman conflict review teaches the organisation model what to trust.\nMemory conflict review\nHow scoped memory works for teams\nCompany context report"},{"type":"technical-note","title":"How scoped memory works for teams","description":"A technical note on personal, team, project, org, and sensitive memory scopes for AI agents at work.","path":"/writing/scoped-memory-for-teams","url":"https://ambience.sh/writing/scoped-memory-for-teams","markdownPath":"/writing/scoped-memory-for-teams.md","updated":"2026-06-07","summary":"Scoped memory turns relevance into policy: this person, this agent, and this project must be allowed to see the context.","sections":[{"title":"The five-scope model","body":["Ambience supports personal, team, project, org, and sensitive memory scopes."]},{"title":"Access levels","body":["Human-readable access levels map to memory actions and audit."]},{"title":"How teams update memory","body":["Teams can amend, rescope, and resolve memory without quiet rewrites."]}],"related":[{"label":"Security and permissions","href":"/company-context/security-and-permissions"},{"label":"Redaction before storage","href":"/writing/redaction-before-storage"},{"label":"How do teams share AI agent memory?","href":"/answers/how-do-teams-share-ai-agent-memory"}],"searchText":"How scoped memory works for teams\nA technical note on personal, team, project, org, and sensitive memory scopes for AI agents at work.\nScoped memory turns relevance into policy: this person, this agent, and this project must be allowed to see the context.\nThe five-scope model\nAmbience supports personal, team, project, org, and sensitive memory scopes.\nAccess levels\nHuman-readable access levels map to memory actions and audit.\nHow teams update memory\nTeams can amend, rescope, and resolve memory without quiet rewrites.\nSecurity and permissions\nRedaction before storage\nHow do teams share AI agent memory?"},{"type":"technical-note","title":"Why redaction has to happen before storage","description":"A technical note on redacting agent context before it becomes durable shared memory.","path":"/writing/redaction-before-storage","url":"https://ambience.sh/writing/redaction-before-storage","markdownPath":"/writing/redaction-before-storage.md","updated":"2026-06-07","summary":"The safer boundary is before persistence: raw sensitive text should not become the durable artifact.","sections":[{"title":"Write path","body":["Candidate memories are proposed, detected, stripped, and audited before durable storage."]},{"title":"Controls","body":["Secrets and PII are removed before durable writes."]},{"title":"Audit evidence","body":["Audit events record policy metadata without exposing the secret."]}],"related":[{"label":"Security and trust","href":"/security"},{"label":"Scoped memory for teams","href":"/writing/scoped-memory-for-teams"},{"label":"How to audit AI agent context","href":"/answers/how-to-audit-ai-agent-context"}],"searchText":"Why redaction has to happen before storage\nA technical note on redacting agent context before it becomes durable shared memory.\nThe safer boundary is before persistence: raw sensitive text should not become the durable artifact.\nWrite path\nCandidate memories are proposed, detected, stripped, and audited before durable storage.\nControls\nSecrets and PII are removed before durable writes.\nAudit evidence\nAudit events record policy metadata without exposing the secret.\nSecurity and trust\nScoped memory for teams\nHow to audit AI agent context"}]}