Langfuse integration for Rig
The easiest way to connect Langfuse to Claude and other AI tools
Langfuse is an open-source LLM observability platform that traces model calls, scores outputs and tracks token usage, cost and latency.
What Rig syncs from Langfuse
Once connected, Rig models these objects so questions resolve to the right tables and joins:
- Traces
- Observations and spans
- Generations
- Scores
- Prompts
- Token usage
- Costs
- Latency
How Langfuse connects to Rig
Rig Ingest syncs your Langfuse traces, generations and scores into your warehouse on a schedule. Rig then models traces, prompts and token usage so cost, latency and quality questions resolve correctly.
Analyse token cost and LLM spend
Token cost is the dollar spend of model usage across prompt and completion tokens. Rig reads Langfuse generations and cost data, so you can break spend down by model, prompt, feature and time, and see what is driving your LLM bill without exporting Langfuse dashboards.
- What is total LLM token cost by model this month?
- Which feature drives the most spend per request?
- Show daily token cost trend over the last 30 days
Track latency and error rates by feature
Rig reads Langfuse trace timings and statuses, so you can measure p50 and p95 latency and error rates per feature and model. Engineering teams see where users wait or calls fail before it shows up in complaints.
- What is p95 latency by feature this week?
- Which model has the highest error rate?
- Show latency trend for the chat endpoint over 14 days
Measure prompt and version quality scores
By joining Langfuse scores to prompts and versions, Rig shows how output quality changes as you iterate. Teams see whether a new prompt version actually improved evaluation scores across traces.
- How did quality scores change after the latest prompt version?
- Which prompt version has the highest average score?
- Show score distribution by feature this month
Analyse trace volume and usage by feature
Rig rolls up Langfuse traces by feature, model and user, so you can see request volume and how usage is growing. Product teams see which AI features are gaining traction and which models carry the load.
- How many traces did each feature generate this month?
- Which AI feature is growing fastest in usage?
- Show trace volume by model over the last 90 days
Ask your Langfuse data in Claude, ChatGPT or Cursor
Rig serves your Langfuse data to AI assistants over MCP. Once Langfuse is synced into your warehouse, you can ask questions in plain English from the tool you already use, and every answer is backed by validated, governed SQL.
Frequently asked questions
- What Langfuse data does Rig sync?
- Traces, observations and spans, generations, scores, prompts, token usage, costs and latency, with the trace structure preserved.
- Can Rig break down LLM cost and latency by feature?
- Yes. Rig models Langfuse traces and generations in your warehouse, so cost, latency, error-rate and quality questions are answered in plain English.
- Can I query Langfuse data in Claude or ChatGPT?
- Yes. Once Langfuse is synced to your warehouse, Rig serves it to Claude, ChatGPT, Cursor and other AI tools over MCP.
- How often does Langfuse data refresh?
- Rig Ingest syncs on a schedule you set so traces, costs and scores stay current.
Related integrations
Connect Langfuse data to AI tools like Claude
Rig builds a governed context layer over your data so every team, and every AI tool, asks questions and gets answers they can trust.