The Langfuse connector for Muse

See what your LLM app is actually doing.

The Langfuse connector brings LLM observability into chat. Query traces and observations, inspect latencies and errors, manage prompts, and review scores — the debugging questions you’d ask in the Langfuse UI, answered conversationally.

It’s most useful when something feels off: slowest traces last hour, lowest-scored prompts yesterday, where the errors cluster.

What Muse can do with Langfuse

Query traces filtered by time, model, or error
Inspect observations: latency, tokens, and costs
Manage and compare prompt versions
Review scores and evaluation results

Try asking

“Show me the slowest traces from the last hour.”
“Which prompts scored lowest yesterday?”
“Find traces where the model errored this morning.”

Safety notes

Trace inspection is read-only.
Prompt changes confirm before applying — a bad prompt edit affects production.
This is a community-built connector. Review its permissions before connecting.

How to connect Langfuse

Open the Muse app, go to settings, and find connectors. Link your Langfuse account with the standard sign-in — Muse never sees your password. Once linked, the connector is available across your conversations. You can scope it to read-only or disconnect it entirely at any time from the same screen.

Questions

Do I need a Langfuse account with data in it?

Yes, the connector reads your Langfuse project. No traces, nothing to query.

Can Muse change my production prompts?

It can propose and, with your confirmation, apply prompt updates. Nothing changes silently.

Related connectors

Connector status last checked October 13, 2026. The authoritative, current list lives in the Muse app's settings, under connectors. Back to the full directory →