AutoMaintainer is an autonomous, multi-agent AI software engineering platform operating natively inside your GitHub repositories.
Built with LangGraph, FastAPI, Next.js, and powered by Llama 3 (via Groq), the system coordinates specialized AI agents to discover issues, design architecture, implement code changes, open Pull Requests, review diffs, self-correct bugs, and merge verified code.
Warning
AutoMaintainer is currently undergoing a massive architectural transformation from a local prototype to an enterprise-grade Web-First SaaS Platform.
During this major migration phase:
- The platform is currently unstable and active workflows may be temporarily non-operational or experience breaking changes.
- Core systems being actively overhauled include:
- Multi-Tenant SaaS Control Plane: Migrating from single-user local state to organization-scoped data isolation and Row Level Security (RLS).
- Durable Task Queue: Replacing in-memory asyncio task execution with Celery + Redis distributed workers and retry mechanisms.
- GitHub App Authentication: Transitioning from personal access tokens (PAT) to GitHub App installation tokens with automated webhook dispatching.
- Ephemeral Sandbox Architecture: Decoupling execution runners for containerized code execution.
- Pro Monaco WebIDE: Integrating full diff inspection and inline AI assistance.
Please follow our GitHub Discussions and Pinned Epics for real-time progress updates.
Important
AutoMaintainer relies on Supabase (PostgreSQL + Realtime Pub/Sub + Auth) for telemetry, run orchestration, and live agent log streaming.
- Public/Demo Database Status: If the free-tier demo Supabase instance is paused or unreachable, historical run tracking and real-time streaming to the Web UI will be disabled. The UI will show a warning banner indicating that persistence is offline.
- Self-Hosted / Developer Database Setup: If you are running or contributing to AutoMaintainer, you must provision your own Supabase project:
- Create a project at supabase.com.
- Run the complete schema migration script
supabase_schema.sqlin your Supabase SQL Editor to create the required tables (organizations,users,repositories,runs,logs,usage_events,ide_sessions). - Supply your own credentials in
.env(backend) and.env.local(dashboard).
- 5-Agent Hierarchy: Tasks are distributed across specialized agents (Architect, Visionary, Reviewer, Implementer, Maintainer).
- Native GitHub Integration: Agents communicate through real GitHub Issues, PR Comments, and Git Branches.
- Zero-Server Code Intelligence: Powered by GitNexus MCP, allowing agents to semantically navigate your repository and build Code Graphs without sending your code to a third-party server.
- Web IDE & Interactive Terminal: A fully integrated, VS Code-style Web IDE in the browser featuring an interactive PTY terminal connecting directly to the backend.
- Self-Correcting Iteration Loop: If the Maintainer AI rejects a PR, the Implementer AI reads the feedback and pushes a new commit to fix the bug!
- Real-time Observability UI: A sleek, dark-mode React dashboard connected via WebSockets/SSE allows you to monitor the AI Crew as they work in real-time.
- Blazing Fast: Powered by Groq's LPU inference, the entire cycle from Architecture to Merged PR can happen in under 20 seconds.
- Cloud Ready: Production-ready deployment configurations for Vercel, Render, and Docker.
- Node.js (v20+)
- Python (3.11+)
- A Groq API Key
- A GitHub Token or GitHub App credentials
- A Supabase Project (Free Tier supported)
- Redis (optional, required for Celery task workers)
Clone the repository:
git clone https://github.com/PxA-Labs/AutoMaintainer.git
cd AutoMaintainerDatabase Migration:
- Open your Supabase SQL Editor.
- Execute the entire contents of
supabase_schema.sql.
Backend Environment (backend/.env):
GROQ_API_KEY="your_groq_api_key_here"
GITHUB_TOKEN="your_github_token_here"
SUPABASE_URL="https://your-project.supabase.co"
SUPABASE_SERVICE_KEY="your_supabase_service_role_key"
REDIS_URL="redis://localhost:6379/0"Frontend Environment (dashboard/.env.local):
NEXT_PUBLIC_SUPABASE_URL="https://your-project.supabase.co"
NEXT_PUBLIC_SUPABASE_ANON_KEY="your_supabase_anon_key"
NEXT_PUBLIC_BACKEND_URL="http://localhost:8000"cd backend
pip install -r requirements.txt
fastapi dev main.pycd dashboard
npm ci
npm run devOpen http://localhost:3000 in your browser.
AutoMaintainer is configured for distributed cloud deployment:
- Frontend (Vercel): Connect the repository to Vercel with root directory set to
dashboard/. Usesdashboard/vercel.jsonfor static export and security headers. - Backend & Workers (Render): Deploy using the included
render.yamlblueprint to launch the FastAPI control plane, Redis broker, and Celery background workers. - Containerized Deployment (Docker / GHCR): Pull the pre-built multi-arch image:
docker pull ghcr.io/pxa-labs/automaintainer:latest
Curious how it works under the hood? Read our Architecture Documentation to see how the 5-agent LangGraph topology operates.
Want to add a new Agent or improve the dashboard? Check out our Contributing Guidelines.
AutoMaintainer maintains a living changelog and automated release notes adhering to Keep a Changelog and Conventional Commits. Review all historical and unreleased updates in our CHANGELOG.md.
This project is licensed under the MIT License.