kaibo
activeEngineering doctrine for coding agents: how your team builds things, kept in one place, reviewed before it counts, and loaded by agents when they start the work.
Why it exists
Every team has a way it builds things: how tests are tiered, what a base image is pinned to, what a pull request has to show. Most of it lives in a few heads, or is pasted into every repository's AGENTS.md, where the copies drift apart. Agents never see the version that was actually agreed on, so their work is plausible, locally consistent, and quietly not how the team does things.
kaibo keeps that doctrine in one place, reviewed before it counts.
Who it's for
- The lead or CTO who wants one standard for how the organisation builds: owned per domain and reviewed, instead of rediscovered repository by repository.
- The engineer tired of copying conventions into every AGENTS.md: write them down once, and every agent session can read, cite and extend them.
How agents use it
- Load it when the work starts. Before an agent begins on, say, testing or observability, it loads the team's doctrine for that domain.
- Research it before deciding. When an agent has to choose, it looks up what the team has already settled and cites it.
- Say so when nothing is written down. kaibo reports a gap instead of letting the agent guess.
- Send additions to the team. What an agent proposes to add goes to the team for review, and a person decides whether it lands.
Under the hood
A single Rust command-line tool that searches on your own machine and answers with citations. There is no model inside it, so the agent does the reading and kaibo stays predictable.
Where it stands
Early release, used internally. Nicolas Bissig and I build it together.