Skip to content
OpenSourceSoftware.ai
LibreChat logo

LibreChat

LibreChat-AI · AI assistants

GitHub stars

46,385
—Awaiting first sync

Awaiting first sync

No measured star history yet

What is LibreChat?

LibreChat puts the familiar AI chat experience on infrastructure you control and removes the lock-in to a single provider. Teams pick the models that suit each task, including local ones, and extend the chat with agents and tools.

LibreChat is an open-source AI chat platform with a ChatGPT-style interface that works with OpenAI, Anthropic, Google, AWS Bedrock, Azure, local models through Ollama and any OpenAI-compatible API. It adds agents, MCP tools, a sandboxed code interpreter and web search, and can be self-hosted.

Core capabilities

  • Chat with OpenAI, Anthropic, Google, Bedrock, local Ollama models and any OpenAI-compatible API.
  • Build no-code agents with MCP tools, file search, code execution and reusable skills.
  • Run code safely in a sandboxed interpreter and search the web from the conversation.

One interface for many models

LibreChat supports Anthropic, AWS Bedrock, OpenAI, Azure OpenAI, Google and Vertex AI, and any OpenAI-compatible endpoint without a proxy. Local and remote providers such as Ollama, Mistral AI, Groq, OpenRouter and DeepSeek work too.

  • Multi-model
  • Ollama
  • OpenAI-compatible

Agents, tools and code execution

No-code agents can use MCP servers, file search, code execution and reusable skills, and can be shared with specific users and groups. A sandboxed code interpreter runs Python, JavaScript, Go and other languages, and web search brings current information into the context.

  • Agents
  • MCP
  • Code interpreter
  • Web search

Built for self-hosting

LibreChat is MIT-licensed and designed to be deployed on your own infrastructure, with OpenID and OAuth sign-in, a trace viewer for model calls and costs, and OpenTelemetry export for logs.

  • Self-hosted
  • MIT
  • Observability

Where it fits

Use cases

  1. 01

    Give a team one AI workspace for every model

    Connect the providers your team uses, plus local models, so people can switch models in one familiar chat interface instead of paying for separate subscriptions.

  2. 02

    Build assistants for a client

    Create no-code agents with their own instructions, MCP tools and files, and share them with specific users or groups.

  3. 03

    Analyse files with code

    Upload a spreadsheet or dataset and let the sandboxed code interpreter process it in Python or another language, then download the results.