OpenZessAutonomous Agent Matrix

Build, deploy, and orchestrate AI agents with native tool access, multi-provider LLM routing, and a sandboxed Linux environment.

Get Started →

View on GitHub

🧠

Multi-Provider LLM Engine

Route prompts to Gemini, OpenAI, Anthropic, Groq, DeepSeek, Qwen, GLM, Kimi, or local Ollama models — all through a single unified interface powered by LiteLLM.

🛡️

Sandboxed Execution

All terminal commands execute inside an isolated Debian WSL sandbox. Your host Windows machine is never exposed. Safety-first by design.

💻

Matrix Viewer

Watch your AI agent interact with an invisible X11 Linux desktop in real-time via WebSocket streaming. Click, type, and browse — all inside the matrix.

🔌

MCP Plugin Ecosystem

Connect to any Model Context Protocol server (Stitch, Prisma, custom tools) via stdio or HTTP transport. Hot-plug external capabilities into your agent's brain.

🌐

Cloud-Native Database

Powered by Neon Serverless PostgreSQL. All sessions, messages, personas, and notes persist safely in the cloud with zero local file risk.

⚔️

Swarm Intelligence

Deploy multiple AI agents simultaneously in the War Room. Watch them debate, critique, and synthesize solutions collaboratively in real-time.

Why OpenZess?

OpenZess is not just another chatbot UI. It is a full operating system for AI agents — a complete environment where autonomous agents can execute code, browse the web, manage files, and even control a virtual desktop.

Built by Debjit Das, OpenZess combines the power of modern LLMs with native system-level capabilities that no other open-source project offers.

What makes it different?

Feature	OpenZess	Typical Chat UIs
Terminal execution	✅ Sandboxed WSL	❌ None
File system access	✅ Full CRUD	❌ None
Multi-provider LLM	✅ 10+ providers	❌ Single provider
Virtual desktop control	✅ X11 Matrix	❌ None
Plugin system	✅ MCP + Python	❌ None
Multi-agent swarm	✅ War Room	❌ None
Cloud database	✅ Neon PostgreSQL	❌ Local only
Telegram/Discord bots	✅ Built-in	❌ None