danny-avila/librechat

LibreChat is a self-hosted, open-source chat interface that works as a unified front-end for dozens of AI language model providers. The problem it solves is that people who work with multiple AI models, ChatGPT, Claude, Gemini, Mistral, DeepSeek, and many others, normally have to maintain separate subscriptions and switch between different websites or apps, each with their own interface. LibreChat provides one consistent chat interface that connects to all of them using your own API keys, running on your own server. In practice, you install LibreChat (typically via Docker), configure your API keys from providers like OpenAI, Anthropic, Google, or even locally running models via Ollama, and then your users get a ChatGPT-style interface that can switch between any of those models mid-conversation. Beyond basic chat, it supports AI Agents, customizable automated assistants you build without code that can use tools, search the web, run code, and access files. Code execution runs in a sandboxed environment supporting Python, Node.js, Go, Rust, and other languages. It also supports the Model Context Protocol (MCP), which lets agents use external tool servers. Conversation search, branching (forking a conversation to explore different directions), image generation, speech-to-text, and multi-user authentication with OAuth2 and LDAP round out the feature set. You would use LibreChat if you want a privacy-respecting, self-hosted alternative to ChatGPT Plus that works with your own API keys across multiple providers, or if you are running an internal AI assistant for a team. The tech stack is TypeScript with a React frontend, Node.js backend, MongoDB for storage, and Docker for deployment.