# codekit

Remote control for AI coding agents, version 2.9.0, answering at https://codekit.mikebild.dev.

A task is one prompt against one project, run by a coding agent in its own git
worktree. The agent plans first, a human approves the plan, and only then does
anything get written.

Everything below is served by this process out of its own binary.

- [/docker-compose.yml](https://codekit.mikebild.dev/docker-compose.yml) — The topology that runs this exact release
- [/install](https://codekit.mikebild.dev/install) — What there is to install on a workstation
- [/install/skill](https://codekit.mikebild.dev/install/skill) — Installer for an agent skill that asks who approves what an agent changes
- [/llms.txt](https://codekit.mikebild.dev/llms.txt) — Should you use codekit, what it does, how to run one — indexed
- [/llms-full.txt](https://codekit.mikebild.dev/llms-full.txt) — Every page it carries, in one fetch, for a model
- [/agent-guide.md](https://codekit.mikebild.dev/agent-guide.md) — How an agent should drive codekit
- [/openapi.json](https://codekit.mikebild.dev/openapi.json) — The full API surface
- [/docs](https://codekit.mikebild.dev/docs) — That API surface as a browsable reference
- [/cockpit](https://codekit.mikebild.dev/cockpit) — The web console
- [/health](https://codekit.mikebild.dev/health) — Liveness
- [/ready](https://codekit.mikebild.dev/ready) — Readiness, with its checks
- [/.well-known/service-descriptor.json](https://codekit.mikebild.dev/.well-known/service-descriptor.json) — All of the above, listed with checksums

Start with https://codekit.mikebild.dev/llms.txt if you are a model, https://codekit.mikebild.dev/cockpit if you are a
person, and https://codekit.mikebild.dev/docker-compose.yml if you want to run one of these
yourself.

This is an index of what the binary serves, not the complete documentation of the product.