AI Alpha Squad — business requests in, working software out.
The first public release of an autonomous multi-agent software delivery squad: open a GitHub Issue, and a team of specialized AI agents analyzes, designs, builds, tests, security-reviews, documents, and opens a pull request — with a human Director approving the milestones that matter.
🌐 Live: aialphasquad.com · 🖥️ Director Dashboard: aialphasquad.com/director
Highlights
- Autonomous end-to-end pipeline across 8 specialized agents (Business Owner → Architect → Developer → QA · Security · DevOps · Tech Writer → Release Manager), orchestrated on GitHub Actions.
- GitHub-native — Issues are the work queue and the audit log; code ships to the target repo via pull requests.
- Open-model inference via Hugging Face (e.g. DeepSeek, Qwen) with automatic escalation to stronger models when QA pushes back.
- Director Dashboard (shadcn/ui) — live lifecycle timeline with timestamps, the squad working in real time, per-agent AI-model history, and Stop / Retry controls.
- Human-governed — one Director, two approval gates, full veto. Optional WhatsApp approvals, mirrored to the issue.
- Open & self-hostable — GitHub Actions + Hugging Face + Cloudflare Workers. MIT licensed.
Status
Experimental research lab. It already ships real pull requests to real repos — and it has rough edges. 0.x reflects that: expect change before a stable 1.0.
Get started: see the README · Brief the squad: open a request.