LibreChat

LibreChat is an open-source AI chat platform from Danny Avila that mirrors the ChatGPT interface while adding multi-provider support, plugins, MCP client integration, and multi-user authentication. The project is commonly deployed as a chat front-end to LLM gateways or directly against OpenAI-compatible endpoints.

What it does

Users sign in through a web UI, choose from the providers the admin has configured, and chat. LibreChat supports OpenAI, Anthropic, Gemini, Azure, AWS Bedrock, Mistral, Ollama, and any OpenAI-compatible endpoint. Conversations can be forked and branched — each branch represents an alternative response path — and compared side by side.

File upload supports vision models for image inputs and document RAG for PDF, Word, and text documents. Conversation presets package a system prompt, model, and parameter set together for reuse. Shared prompts let teams publish templates across users.

Authentication covers email-password, OAuth (GitHub, Google, Discord), and generic OpenID Connect. The plugin system provides web search, code execution, and custom tools. MCP client support lets conversations invoke MCP tools inline.

Licensing

MIT. Development runs on a donations-supported community model.

Deployment

Docker Compose reference stack with MongoDB and optional MeiliSearch for conversation search. Runs on Docker, Docker Swarm, and Kubernetes.

Limitations

• MongoDB is the required primary database; there is no Postgres backend, which may not fit stacks standardized on Postgres.
• Admin configuration is driven by environment variables and JSON files rather than an admin UI — most settings are edited on disk.
• Feature scope is chat-UI-first; for programmatic clients like Cursor or Claude Code, LibreChat is not the endpoint — a gateway sits upstream.
• Plugin authoring uses LibreChat's specific plugin format; community plugins exist but the catalog is smaller than hosted alternatives.