About team.management.
team.management is an open-source protocol engine for Claude Code. It puts your development process on rails — the steps, gates, and reviews run as the AI works, and you get a record of what actually happened, not a claim that it did.
What it is
A native Claude Code plugin. You describe your process as a protocol — just steps and gates — and the engine holds the AI to it: hooks block tool use until the current step allows it (the DAIC loop), a spec-review gate stops the agent advancing until the diff matches the spec, and every run leaves an audit trail. It is free and open source under the MIT license: no accounts, no API keys, no paid tier.
Who makes it
An independent project, conceived and directed by Alex Rozanski and implemented by Max Tushkov. It is not affiliated with or endorsed by any AI provider; all product names and logos are the property of their respective owners. The source lives on GitHub.
Frequently asked
What is team.management?
An open-source protocol engine for Claude Code. It puts your development process on rails — the steps, gates, and reviews are enforced as the AI works, and you get a record of what actually happened.
Is it free?
Yes. team.management is open source under the MIT license — no accounts, no API keys, no paid tier, no signup.
How do I install it?
It is a native Claude Code plugin. Add the marketplace, install the plugin, and run init — your team follows via one committed file. The install guide is at team.management/docs/install.
Does it enforce the process, or just suggest it?
It enforces. Hooks block tool use until the current step allows it (the DAIC loop), and a spec-review gate stops the agent advancing until the diff matches the spec.
Does it work with agents other than Claude?
It is built for Claude Code. Codex and Antigravity can join as parallel collaborators across several workflow phases when you enable them.
What does team.management guarantee?
It attests that your defined process ran — the steps, gates, and reviews. It does not warrant that the output is correct: correctness depends on your task specification and the tests you set. A well-followed process on a poorly-specified task still produces the wrong thing.
How do I get help or report a problem?
Through GitHub issues on the plugin repository. There is no signup and no support inbox — the project is developed in the open.