brainrc¶
brainrc installs one set of AI coding-agent config into Claude Code, Codex CLI, OpenCode,
Antigravity CLI and Cursor. You keep your rules, skills, agents, guard hooks and MCP servers
as plain files in git, and brainrc apply writes each piece where each tool looks for it.
uvx brainrc status # preview: lists what apply would write, writes nothing
uv tool install brainrc # or: pipx install brainrc
brainrc apply # write it
What you get¶
- Rules every session loads, rendered into each tool's instructions file (
~/.claude/CLAUDE.md,~/.codex/AGENTS.md, ...). - Guard hooks that ask before risky commands (
git push,terraform apply,kubectl delete, ...) and block secrets from being written (guardrails). - Gatework, the default pack: skills and agents that take a ticket to a reviewed pull request, with you approving the spec, the plan and the push (Gatework).
- Your own packs: a folder or git repo of your rules and skills, layered on top (packs).
- A lockfile that pins every pack to a commit and every file to a hash, so a team's repo gets the same agent files on every machine and in CI (team setup).
brainrc never overwrites a file it didn't write. It merges into your settings files and keeps a backup.
Where to start¶
| You want to | Read |
|---|---|
| Install it | Install |
| Try it end to end in ten minutes | Getting started |
| Learn the words (pack, harness, profile, lockfile) | Concepts |
| Add your own rules and skills | Packs, then your first pack |
| Give a repo's agents the same setup on every machine | Team setup |
| Look up a command or a config file | Commands, configuration |
| Fix something | Troubleshooting |
The source is github.com/brainrc/brainrc. Each release has its own version of this site; the version menu at the top switches between them.