← use cases · governance
Auditable, reproducible agent runs
Keep an immutable record of every task, session, lock, and merge for review and reproducibility.
who it's for
Teams that need to answer "what did the agents actually change, and why" — for review, compliance, or debugging a bad run.
the problem
When agents edit your repo autonomously, "it worked on my machine" is not good enough. You need a durable, tamper-evident record of every decision the harness made.
how brat does it
- 1
Every state change — task created, session started, lock acquired, result merged — is written as an immutable event to refs/grite/wal before it counts.
- 2
The log lives in your git refs, so it travels with the repo and needs no external service.
- 3
Replay the WAL to rebuild the exact materialized view at any point in history.
- 4
Use the log to reconstruct precisely which agent touched which files, when, and under which task.
the outcome
A complete, replayable audit trail of an autonomous agent fleet — the accountability layer that raw parallel execution never provides.