Langfuse Prompt Management MCP Server

What is Langfuse Prompt Management MCP Server?

A Model Context Protocol (MCP) server that integrates with Langfuse Prompt Management, enabling LLM clients like Claude Desktop to discover, retrieve, and compile Langfuse prompts through the MCP interface. It is designed for developers using Langfuse to manage prompt versions.

How to use Langfuse Prompt Management MCP Server?

Install dependencies (npm install), build the server (npm run build), then add it to your MCP client’s configuration (e.g., claude_desktop_config.json or Cursor) with your Langfuse API keys and base URL set as environment variables. The server exposes both standard MCP prompts and equivalent tools for broader client compatibility.

Key features of Langfuse Prompt Management MCP Server

• Implements MCP Prompts specification (prompts/list, prompts/get)
• Also provides tools (get-prompts, get-prompt) for non-prompt-capable clients
• Transforms Langfuse text and chat prompts into MCP prompt objects
• Compiles prompts with provided variables
• Supports optional cursor-based pagination for listing prompts

Use cases of Langfuse Prompt Management MCP Server

• Retrieve and compile production prompts from Langfuse inside AI chat assistants
• Manage and iterate prompt versions without modifying client code
• Integrate Langfuse prompt management with MCP-compatible LLM tools like Claude Desktop

FAQ from Langfuse Prompt Management MCP Server

What prompts are accessible?

Only prompts with the production label in Langfuse are returned.

What are the runtime dependencies?

Node.js is required. You need a Langfuse account with a public and secret API key.

Can I use a self-hosted Langfuse?

Yes, by setting the LANGFUSE_BASEURL environment variable to your self-hosted instance.

How does authentication work?

Authentication uses the LANGFUSE_PUBLIC_KEY and LANGFUSE_SECRET_KEY environment variables.

Does the server support chat prompts?

Yes, both Langfuse text and chat prompts are transformed into MCP prompt objects.