Skip to main content
Three commands stand between a fresh install and a tracked, test-driven coding loop:
Then, inside your AI tool:
That’s the whole setup. Here’s what each step actually did:
1

Install

Puts the tiny-brain CLI on your path. See Install for Homebrew, the installer script, and the Claude Desktop bundle.
2

tiny-brain init

Analyses your repo, installs the TDD git hooks, and writes the agent context. From here on, commits follow the RED → GREEN → REFACTOR pipeline and are validated by the hooks.
3

/plan

Your agent works with you to shape the request into a PRD — features and tasks as markdown in docs/prd/ — then works through the tasks test-first, with an adversarial review after every implementation commit.
While the loop runs, the dashboard streams every step at http://localhost:8765 — plans in flight, per-task Red/Green/Review pipelines, diffs, and review verdicts. Take the dashboard tour to see what’s there. To understand what the agent is doing on your behalf — and what the hooks will and won’t let it commit — walk through your first PRD.