Rusuban docs
Everything about running Rusuban: install it, write the rules its agents follow, and know what holds them.
- Install
- Install Rusuban with npm: Node 22 or later, gh and Claude Code logged in. Then rusuban init sets up the service, your first repo and the dashboard.
- How it works
- Rusuban reads your repos, has Claude Code judge each issue and PR against your rules, lets agents act in worktrees, merges behind locks, and asks you the rest.
- Writing rules
- Rusuban's agents follow plain-text rules in your repo's .rusuban/rules/: where they live, how each run gets them, what to write, and how to try them.
- Locks
- The eight gates Rusuban's server checks on live GitHub state before it merges a pull request, the built-in locks, merge orders and a repo's locks.json.
- Security model
- What holds Claude Code agents that run on your machine with your logins: deny rules, the host profile, merges by the server only, and the risks that remain.
- FAQ
- Rusuban's rules, merges, your machine and your accounts: whose Claude plan the agents use, who they act as on GitHub, and how to stop everything.
- Set up from Claude
- Rusuban's MCP server lets Claude Code read your instance and change its settings. A change that widens what agents may do waits for your yes.
- Test machines
- Give Rusuban's agents Linux, macOS or Windows machines over SSH, reserved one agent at a time, with a setup script for each system.