Open-Antigravity is an ambitious open-source project aiming to create a universal AI gateway and agent-first IDE, serving as a model-agnostic alternative to proprietary platforms. It targets developers seeking true model freedom, democratized AI access, transparency, and self-hosting capabilities, enabling them to delegate complex tasks to autonomous AI agents powered by their chosen LLMs.

How It Works

This project is architected as a modular, container-native application. A React-based web UI provides both a VS Code-based editor and a dedicated manager view for orchestrating AI agents. The core logic resides in a Gateway & Orchestrator (FastAPI/Node.js) that interprets requests, dispatches tasks, and manages workflows. It leverages a Workspace Manager for isolated, containerized agent environments and an AI Model Gateway, which acts as a standardized interface to connect with diverse LLMs, ensuring model agnosticism and flexibility.

Quick Start & Requirements

• Primary install/run command: Clone the repository, set up .env with API keys, and run docker-compose up --build.
• Prerequisites: Docker and Docker Compose, Node.js (v20+), Python (v3.11+).
• Access: The platform is accessible at http://localhost:3000.
• Links: ROADMAP.md, Contribution Guide, Discord server.

Highlighted Details

• Universal LLM Gateway: Designed to connect seamlessly with a wide array of LLMs, including proprietary (OpenAI, Anthropic, Gemini) and open-source models (Llama, Grok), eliminating vendor lock-in.
• Agent-First Workflow & Multi-Agent Collaboration: Enables delegation of high-level tasks to autonomous AI agents, supporting parallel execution and collaboration among multiple agents for complex development goals.
• Dual-View Interface: Combines a feature-rich, AI-enhanced IDE (based on VSCodium) with a dedicated Manager View for spawning, monitoring, and orchestrating AI agents.
• Trust Through Verifiable Artifacts: Agents produce tangible outputs like task plans, screenshots, recordings, and test results, allowing users to verify autonomous work.
• Interactive Feedback Loop: Facilitates real-time user feedback on agent actions and artifacts, enabling agents to incorporate corrections and guidance without losing context.

Maintenance & Community

The project actively encourages community contributions via a detailed Contribution Guide and a Discord server for developer interaction. A living roadmap is maintained in ROADMAP.md.

Licensing & Compatibility

This project is licensed under the MIT License, which is permissive for commercial use and integration into closed-source projects.

Limitations & Caveats

The project explicitly states it is "not intended for production use" and is an independent initiative, not affiliated with Google.