Build, run, and share AI agents that actually finish work

LambChat is an open-source AI Agent platform: multi-model chat, MCP tools, a skills engine, persistent memory, and production-ready deployment β€” in one self-hostable stack.

LambChat web interface showing an AI agent conversation with streaming output, tool calls, and skill panels

What LambChat can do

πŸ€– Agent Runtime

Deep agent graphs with sub-agents, thinking mode, streaming output, and human approval for risky actions.

πŸ”§ MCP & Tools

System-level and per-user MCP servers, encrypted secrets, and sandboxed code execution on Daytona, E2B, or CubeSandbox.

🧠 Memory & Skills

Cross-session memory, a skill marketplace, GitHub-synced skills, and reusable persona presets.

πŸ’¬ Multi-Model Chat

Bring your own keys for Claude, GPT, Gemini, and more β€” with SSE streaming, multimodal input, and document processing.

πŸ“± Full-Stack Client

React 19 web app, Capacitor mobile apps, Tauri desktop app, and installable PWA β€” one backend for all of them.

πŸš€ Production Ready

JWT auth with RBAC, realtime sync, scheduled tasks, usage tracking, and Docker / Kubernetes deployment recipes.

Why LambChat

Most agent products stop at β€œchat with tools.” LambChat is designed for the longer path: configure models, connect tools safely, let agents create artifacts, persist useful context, share results, approve risky actions, and deploy the whole system for real users β€” self-hosted, open source (MIT), and bilingual documentation included.

Under the hood

Python 3.12 + FastAPI + LangGraph / deepagents Β· MongoDB + Redis + arq Β· React 19 + TypeScript + Vite Β· Deploy with Docker or Kubernetes. English, δΈ­ζ–‡, ζ—₯本θͺž, ν•œκ΅­μ–΄, and Русский interfaces.