// the find
LibreChat-AI/LibreChat
Enhanced ChatGPT Clone: Features Agents, MCP, Skills, DeepSeek, Anthropic, AWS, OpenAI, Responses API, Azure, Groq, o1, GPT-5, Mistral, OpenRouter, Vertex AI, Gemini, Artifacts, AI model switching, message search, Code Interpreter, langchain, DALL-E-3, OpenAPI Actions, Functions, Secure Multi-User Auth, Presets, open-source for self-hosting. Active
LibreChat is a self-hosted, multi-provider chat UI that's grown into a full agent platform — MCP tools, code execution, RAG, image generation, and now subagents and 'skills' sit on top of the original ChatGPT-clone core. It's for teams or individuals who want one interface in front of every model provider and are willing to run the infrastructure to get it.
The provider abstraction is real engineering, not config flags — switching between Anthropic, Bedrock, Azure, Vertex, and arbitrary OpenAI-compatible endpoints doesn't touch app code. Resumable streams over Redis mean a dropped connection doesn't kill an in-flight generation, a detail most self-hosted chat UIs never bother with. Code Interpreter runs in an isolated sandbox rather than shelling out on the host, which is the right default for anything user-facing. The admin panel and RBAC ship inside the Docker Compose stack itself instead of being a separate bolt-on tool.
The feature surface has exploded — agents, subagents, skills, attached workspaces, trace viewer, MCP, artifacts — and the changelog ships breaking changes almost every release, which is a lot to keep stable at once. A full deployment needs MongoDB, Redis, and Meilisearch minimum, plus a vector store if you want RAG, before a single message is sent. Several headline features in the current release (attached workspaces, subagents) are labeled 'highly experimental' right alongside stable ones, and the docs don't draw a hard line between the two. MongoDB as the datastore for something this relational — users, conversations, agents, shared links, permissions — shows up as recurring consistency and migration friction in the issue tracker.