Creativity · MCP — integration

MCP Integration: LibreChat

LibreChat is a popular open-source chat app that proxies many LLM providers behind one UI. Since 2025 it supports MCP servers as first-class tool providers, so you can configure a Postgres MCP and a Slack MCP and use them with OpenAI, Anthropic, and Ollama in the same LibreChat install.

MCP facts

Kind: integration
Ecosystem: anthropic-mcp
Language: TypeScript / Node.js
Transports: stdio, http

Capabilities

Consumes MCP servers across OpenAI, Anthropic, Google, and local models
Configure servers declaratively in librechat.yaml
Per-user and per-endpoint tool selection

Install

docker compose up -d  # from the LibreChat repo

Configuration

# librechat.yaml
mcpServers:
  filesystem:
    command: npx
    args: ["-y", "@modelcontextprotocol/server-filesystem", "/workspace"]
  github:
    command: npx
    args: ["-y", "@modelcontextprotocol/server-github"]
    env:
      GITHUB_TOKEN: "${GITHUB_TOKEN}"

Frequently asked questions

How does LibreChat's MCP differ from Claude Desktop's?

Same protocol, different client surface. LibreChat runs server-side and is multi-user/multi-provider, while Claude Desktop is a single-user desktop client.

Does it work with all providers?

Tool calling works with any provider whose model supports function calling — OpenAI, Anthropic, Gemini, Mistral, and many open-source models through Ollama.

Safety concerns?

Server-side MCP means tokens live on the server, shared across users unless scoped. Use per-user credentials where possible and follow the LibreChat docs on MCP isolation.

Sources

LibreChat Documentation — accessed 2026-04-20
Model Context Protocol — accessed 2026-04-20