LibreChat-AI/LibreChat

What LibreChat does

LibreChat is a self-hosted AI chat platform that puts the major model providers behind one ChatGPT-like interface: OpenAI, Anthropic, Azure OpenAI, AWS Bedrock, Google and Vertex AI, plus any OpenAI-compatible custom endpoint, so local Ollama or other providers can also be wired in. It is aimed at individuals and organizations that want control over their AI infrastructure rather than a single vendor's hosted client.

Beyond plain conversation, it adds Agents (no-code assistants with sharing and an Agent Marketplace), Model Context Protocol tools, reusable SKILL.md instruction bundles, subagents, code artifacts, sandboxed code execution in several languages, web and file search, image generation, speech input/output, and import/export of conversations. Multi-user authentication (OAuth2, LDAP, email), a browser-based admin panel and token-spend controls make it usable as a shared internal service.

Getting started

The project ships Docker Compose stacks that bundle the admin panel, so a self-hosted instance is the standard path. The README links one-command deployment options on Railway, Zeabur and Sealos for a quick trial; for local or production installs, follow the documentation rather than improvising configuration, since the concrete .env contents are deployment-specific and not spelled out in the README.

Once the stack is running, open the UI, sign in (set up an admin account as described in the docs), and configure model endpoints and API keys through the configuration files or the admin panel. At that point you can create an Agent, attach MCP servers or Skills, and switch between endpoints and presets mid-chat.

If you only want a hosted trial before committing to self-hosting, the Railway deploy button is the shortest route, but it still requires provider API keys for whichever models you want to use.

Constraints

• LibreChat is a self-hosted application: you need to run the server and database (and Redis for horizontally scaled or resumable-stream deployments) yourself, and you are responsible for upgrades.
• The README explicitly warns to check the changelog for breaking changes before updating; the project is in active development and features are marked beta or highly experimental (Agent Management API, attached code workspaces).
• Some features depend on external services and keys: TTS/STT via OpenAI, Azure or Elevenlabs; file search via the separate RAG API; image generation via provider APIs or MCP servers; web search via configured providers.
• Code execution runs in an isolated sandbox powered by ClickHouse/code-interpreter, but code approval settings (Ask/Allow/Deny, including a Full access mode) should be configured carefully in shared deployments.
• Licensed under MIT, so you can modify and redistribute it, but you supply the infrastructure and provider accounts.

HotRepo's take

On the TypeScript daily list for 2026-11-01, LibreChat recorded 61 new stars and a best position of 13 on that list. That places it near the top of a language-specific daily ranking rather than the all-language one, and the snapshot is a single day, so it supports the observation that the project is still attracting steady attention among TypeScript projects, not a claim about its long-term growth curve. The breadth of the feature surface — multi-provider routing, agents, MCP, admin controls — explains why it competes directly with simpler ChatGPT clones; the trade-off for adopters is that the same breadth makes initial configuration and upgrade tracking more involved than a single-provider client.