Skip to main content
tiny-brain wraps your AI coding loop with guardrails: every piece of work is a tracked task, every task moves through a RED → GREEN → REFACTOR pipeline enforced by git hooks, and every implementation commit is adversarially reviewed before the cycle closes. A local dashboard streams the whole thing live from your machine.

​
What it does

Plans as markdown

PRDs, features, fixes, and tasks live as markdown in your repo. The markdown is the source of truth; progress state is projected from it automatically.

TDD enforced by hooks

Commit-msg and post-commit hooks derive the phase of every task from its commit history — failing tests first, minimal implementation second, review-driven refactor third.

Adversarial review

After every implementation commit, an isolated review agent challenges the work and returns a clean or needs-refactoring verdict.

Local dashboard

Watch plans, phases, reviews, and quality runs stream to a dashboard on your own machine — only your AI tool’s token calls leave it, to a provider you’ve authorised.

​
Install

Install with Homebrew: 
brew install magic-ingredients/tap/tiny-brain
Or with the installer script: 
curl -fsSL https://get.tiny-brain.com | sh
Then bootstrap your repo: 
tiny-brain init

​
Learn more

Quickstart

From zero to a live, guarded loop in three commands.

Your first PRD

Follow one feature through plan, RED, GREEN, and review.

Dashboard tour

What the local dashboard shows while the loop runs.

Releases

Downloads and release notes on GitHub.
The Concepts section explains the machinery — the loop, phase derivation, adversarial review, personas, and the tracking model. The Markdown Format, CLI Reference, and MCP Tools sections cover the surfaces you’ll touch day to day.