Guide preview
Claude Code + Codex in one workspace
How one shared project memory can guide two agents without mixing up the rules for each one.
Public write-up in progress.
Resources
Patterns, explainers, and playbooks.
How other developers wire Munin into their past agent sessions. What changed after the first month.
Replay-check beta
The startup-brief check is in beta now.
We are testing the part that replays old sessions and checks whether the handoff brief still works. If it does not, Munin falls back.
How to import a strategy document
Take a plain strategy memo and turn it into a short list of priorities your agent can actually use.
Guide preview - public write-up in progress.
When should a repeated fix become a rule?
A worked example of when Munin should keep a fix and when it should leave it alone.
Guide preview - public write-up in progress.
Running Munin in a monorepo
How to keep one root memory while still keeping rules scoped to the right package.
Guide preview - public write-up in progress.
Safe handoffs between agents
How to check the handoff brief before you switch tools in the middle of a task.
Guide preview - public write-up in progress.
Writing a guardrail hook
Turn a mistake that keeps repeating into a lightweight check that stops it early.
Guide preview - public write-up in progress.
Debugging stale memory
What to do when the agent keeps pushing an old answer that no longer fits the work.
Guide preview - public write-up in progress.
Changelog
What's new
- v0.4.0 - Apr 2026: Local session import across Claude Code and Codex.
- v0.3.2 - Mar 2026: Replay-check beta. Verification packets shipped.
- v0.3.0 - Feb 2026: Per-agent repeated-mistake reports. Guardrail hooks.
- v0.2.0 - Jan 2026: Rule confirm/decay lifecycle. Freshness & stability scoring.
Want the deeper walkthroughs?
Start with the blog and docs. More implementation notes and public build write-ups will land as the repo opens up.