LibreChat
Open-source chat interface for self-hosting that brings models from many providers and local models together with agents, RAG and SSO.
What it does
LibreChat is an open-source chat interface under the MIT licence that you run on your own servers. You use models from OpenAI, Anthropic, Google, Azure or AWS side by side and can connect local models, for example via Ollama, through custom endpoints. Agents process files, call APIs and use external tools via MCP. On top of that come search in uploaded documents via RAG, code execution in an isolated environment, artifacts, web search, and sign-in via SSO (OAuth, SAML), LDAP and two-factor authentication.
Who it suits
Companies with their own IT that want to give employees a shared AI chat with central user management and decide for themselves which models are used.
Data protection for businesses
Because you host LibreChat yourself, you choose the server location – so running it in the EU is possible. Conversations are not sent to the project, but they do go to whichever model provider you connect. Whether inputs are used for training and which data processing agreement applies is something you settle with that provider. With local models, data stays entirely within your own network. LibreChat itself does not offer a data processing agreement or business plans.
Limits
Setup, updates and security are your responsibility; you need Docker or Kubernetes skills. The project does not mention paid support. The public demo is not suitable for confidential data.