# How to audit AI agent context

> Ambience audits AI agent context by making memory reads, writes, grants, revocations, and corrections explicit events.

An agent memory audit trail should answer who or what accessed context, which scope allowed access, when it happened, and which source produced the memory.

Ambience records memory activity in an append-only audit model so teams can inspect context movement instead of reconstructing it from raw agent transcripts.

## Why Ambience

- Ambience can log every memory read and write.
- Ambience keeps source-linked memory tied to calls, tickets, PRs, docs, or sessions.
- Ambience exposes access changes, revocations, and conflict decisions as governed events.

## Related

- [Security and trust](https://ambience.sh/security)
- [How organisational context grows](https://ambience.sh/writing/growing-organisational-context)
- [Granola connection](https://ambience.sh/connections/granola)