LibreChat is a self-hosted AI chat platform that puts many model providers behind one ChatGPT-style interface. It connects to Anthropic, AWS Bedrock, OpenAI, Azure OpenAI, Google, Vertex AI, and the OpenAI Responses API, and any OpenAI-compatible endpoint can be added as a custom endpoint without a proxy, which covers local and remote providers such as Ollama, Groq, Mistral, Apple MLX, OpenRouter, Deepseek, and Qwen.
Beyond chat, the platform includes LibreChat Agents: no-code assistants that can use MCP servers, tools, file search, and code execution, and that can be shared with specific users and groups or published to a marketplace. Agents support reusable SKILL.md instruction bundles, experimental Agent Plugins that bundle Skills and MCP servers, subagents that run in isolated child contexts, memory with optional per-agent isolation, and human-in-the-loop flows that pause for input or tool approval and can be interrupted or steered mid-run.
A Code Interpreter API runs Python, Node.js, Go, C/C++, Java, PHP, Rust, and Fortran in a sandbox with file upload and download, and code artifacts render React, HTML, and Mermaid content in chat. Deployment covers Docker, reverse proxies, S3 with CloudFront for media, and horizontally scaled setups with Redis, and administration includes OAuth2, LDAP, and email login, a browser-based admin panel for users, groups, roles, and configuration overrides, moderation, and token-spend controls.
Features
- Provider coverage: major cloud providers plus any OpenAI-compatible API as a custom endpoint
- Agents: no-code assistants with MCP, tools, file search, code execution, Skills, Plugins, subagents, and memory
- Human-in-the-loop: agents ask up to four related questions in one form, pause for approval, and can be steered mid-run
- Code Interpreter: sandboxed execution in nine languages with file handling and background runs
- Web search: combines search providers, content scrapers, and rerankers, including configurable Jina reranking
- Artifacts: React, HTML, and Mermaid output with fullscreen preview and SVG/PNG export
- Image generation: GPT-Image-1, DALL-E, Stable Diffusion, Flux, or any MCP server
- Conversation control: presets, mid-chat endpoint switching, branching, forking, import from ChatGPT and Chatbot UI, and export to screenshot, Markdown, text, or JSON
- Resumable streams: responses reconnect after connection drops, with multi-tab and multi-device sync
- Speech: speech-to-text and text-to-speech through OpenAI, Azure OpenAI, and ElevenLabs
- Multi-user administration: OAuth2, LDAP, and email auth, an admin panel, moderation, and token-spend tools
- Observability: encrypted Langfuse connections with per-tenant trace fan-out
