This project provides a framework for developing and deploying meta-agentic AGI systems, aiming to identify, learn, think, design, strategize, and execute across various industries. It targets researchers, developers, and power users interested in advanced AI applications for economic transformation and strategic advantage. The core benefit is a structured approach to building autonomous, self-improving AI agents capable of generating economic value.

How It Works

The system employs a "Meta-Agentic" approach, where agents are designed to create, select, evaluate, and reconfigure other agents and their interaction rules. It utilizes a "Multi-Agent AI DAO" model for scalable infrastructure and incorporates an "Experience-First Loop" (Sense → Imagine → Act → Adapt) powered by techniques like MuZero for latent planning. The architecture includes a central orchestrator, a world model, and a vector-graph memory fabric (pgvector + Neo4j) for knowledge representation. Agents are designed to be polyglot, with graceful degradation for GPU-less or API-key-less operation.

Quick Start & Requirements

• Local Setup: Clone the repository (v0.1.0-alpha tag recommended), install dependencies using pip install -r requirements.lock or pip install -r requirements-dev.txt, and run ./quickstart.sh.
• Docker: Use docker compose up --build or docker run --pull=always -p 8000:8000 ghcr.io/montrealai/alpha-factory:latest.
• Prerequisites: Python 3.11–3.13, Node.js ≥22.17.1 (for browser demos), Docker Compose ≥2.5. Optional: OpenAI API key, Gradio, openai-agents, google-adk, llama-cpp-python or ctransformers for offline inference.
• Resources: Initial dependency installation can take up to 10 minutes. Offline setup requires building a wheelhouse.
• Documentation: Full documentation is available at https://montrealai.github.io/AGI-Alpha-Agent-v0/.

Highlighted Details

• Features a "Gallery Showcase" with 14 demos illustrating various agent capabilities, from financial forecasting to self-healing code repositories.
• Supports both online (OpenAI API, Google ADK) and offline inference modes, with fallback mechanisms for missing dependencies or network access.
• Includes a robust CI/CD pipeline with linting, unit tests, and Docker builds, with manual triggers for owner-initiated workflows.
• Employs a "Zero-Trust Core" with SPIFFE identities, signed artifacts, and audit logs for security and compliance.

Maintenance & Community

The project is actively maintained, with Vincent Boucher credited as a pioneer. Community contributions are welcomed, with detailed guides in AGENTS.md. Links to documentation and contribution guidelines are provided.

Licensing & Compatibility

Distributed under the Apache 2.0 license, allowing for commercial use and integration with closed-source projects.

Limitations & Caveats

The project is described as a "conceptual research prototype" with aspirational goals for AGI and superintelligence. Users are advised to use it at their own risk, and no financial advice is provided. Some demos, like alpha_agi_business_3_v1, are not included in the published package and require running from source.