foundry

<p align="center">
  <img src="docs/hero.png" alt="Foundry — GitHub Issues that write their own code" width="100%">
</p>

<h1 align="center">Foundry</h1>

<p align="center">
  <strong>Multi-agent code factory. GitHub Issues that write their own code.</strong><br>
  One label. Three AI reviewers. Zero human keystrokes.<br>
  Running on a 2019 MacBook Pro for $400/month.
</p>

<p align="center">
  <a href="#quick-start">Quick Start</a> •
  <a href="#how-it-works">How It Works</a> •
  <a href="#the-review-loop">Review Loop</a> •
  <a href="#openclaw-integration">OpenClaw + ACP</a> •
  <a href="#real-numbers">Real Numbers</a>
</p>

---

## The Pitch

Add a `foundry` label to a GitHub Issue. Go to sleep. Wake up to a pull request with three AI code reviews, all fixes applied, CI green.

<p align="center">
  <img src="docs/gifs/foundry-issue-to-pr.gif" alt="GitHub Issue → PR in one command" width="100%">
</p>

No prompts. No terminals. No babysitting. The issue body IS the spec. The agent figures out the rest.

---

## Real Numbers

Early production use across 6 private repos (numbers from internal testing, updated periodically):

| Metric | Value |
|---|---|
| Tasks spawned | 47 |
| Merged successfully | 42 (89%) |
| Average time to merge | 4.2 hours |
| Cost per task | $2-8 |
| Avg review-fix cycles | 3.7 |
| Required human help | 5 (11%) |

Most failures trace back to vague specs, not agent limitations. Fix the spec, re-run, it works.

**Hardware:** 2019 MacBook Pro 16" (Intel i9, 64GB RAM). Two Gemini Max subscriptions ($200/mo each), Codex, Gemini. ~$400/month total.

---

## How It Works

<p align="center">
  <img src="docs/architecture.png" alt="Foundry Architecture" width="100%">
</p>

### The Full Chain

```
GitHub Issue (with `foundry` label)
    ↓
Foundry Orchestrator reads the issue body as a spec
    ↓
Routes to best agent (Claude Code / Codex / Gemini)
    ↓
Creates git worktree + branch
    ↓
Spawns agent with the spec as context
    ↓
Agent writes code, op