Langfuse MCP Server

other

Once your app is running in production, the useful questions shift from "how does Langfuse work" to "what's actually happening in my traces." This connection queries your own Langfuse project — traces, prompts, datasets, and evaluation results — so you can monitor and debug your LLM application's real behavior conversationally instead of digging through the Langfuse dashboard.

• Queries traces, prompts, datasets, and evaluation results from your own Langfuse project
• Built for monitoring and debugging production LLM application behavior
• Lets you investigate a specific trace or prompt version without opening the dashboard

What you can automate

Langfuse MCP Server

Query traces, prompts, datasets, and evaluation results from a Langfuse project to monitor and debug LLM application behavior.

Real workflows

Investigate a bad output

Ask Neotask to pull the trace for a specific request that produced a bad output. It queries your Langfuse project for the trace and associated prompt version.

Check evaluation results after a prompt change

Ask how your latest prompt version scored on evaluation. Neotask pulls the results from Langfuse so you can compare against the prior version.

Frequently asked questions

Is this connected to my own project data?

Yes — it queries your connected Langfuse project's traces, prompts, datasets, and evaluation results.

Can it help me find why a specific run failed?

Yes, ask about a specific trace or time window and Neotask pulls the relevant data to investigate.