Langfuse Skill: Trace, Evaluate and Manage Prompts for LLM Apps with Claude

Teaches Claude to work with Langfuse: add tracing, manage prompts, build datasets and evals, and query your traces through the Langfuse CLI.

Installs: 20.3K; on skills.sh
Stars: 322; on GitHub
Files: 13; besides SKILL.md
Updated: Oct 2026; last repo commit

Langfuse is a skill, published in the langfuse GitHub organization, that makes Claude work with Langfuse, the open-source LLM engineering platform. Claude fetches the current docs first instead of relying on memory, uses langfuse-cli to read and change your project data, and follows task guides (13 reference files in the folder) for tracing, datasets, prompts, evals, user feedback, cost cuts and CI gates.

You need a Langfuse account (cloud or self-hosted) and a project API key pair. SKILL.md tells Claude not to ask you to paste keys into chat. Set them as environment variables instead.

Good for

Developers adding tracing to an LLM app or agent
AI engineers building datasets, evaluators and CI regression checks
Teams moving prompts into Langfuse prompt management
Langfuse users who want Claude to query traces, scores and datasets

Install langfuse

Claude Code

Copy this prompt.

Install the langfuse skill in Claude Code. Run this command:

npx skills add https://github.com/langfuse/skills --skill langfuse

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

From its listing on skills.sh. Source.

What's in the skill

When it loads

Interact with Langfuse and access its documentation: tracing, monitoring, creating datasets, running experiments, and evaluating AI applications. Use when needing to (1) query or modify Langfuse data, (2) look up Langfuse documentation, concepts, integration guides, a feature or SDK usage, or (3) do any AI engineering task (AI observability, prompt engineering/management, evaluation and evaluator management, experimentation, dataset management, evaluation-driven CI/CD, feedback collection). Invoke it for tasks in this scope even when Langfuse is not configured or explicitly mentioned.

Sections in SKILL.md

Core Principles
Use case specific references
1. Langfuse API via CLI
Credentials
Detailed CLI Reference
2. Langfuse Documentation
2a. Documentation Index (llms.txt)
2b. Fetch Individual Pages as Markdown
2c. Search Documentation
Documentation Workflow
Skill Feedback

Key facts

Author: Langfuse
License: MIT
Installs on skills.sh: 20,282 (Nov 2, 2026)
GitHub stars: 322 for the whole langfuse/skills repo (Nov 2, 2026)
Last commit: Oct 27, 2026
What is inside: 14 files including SKILL.md: 13 reference guides, no scripts.
Needs: A Langfuse account and project API keys (LANGFUSE_PUBLIC_KEY, LANGFUSE_SECRET_KEY, LANGFUSE_BASE_URL)
Also in the repo: migrate-to-langfuse, a separate skill for moving from another observability platform
Works in: Claude Code, Cursor and other assistants (README); Cursor plugin available

What it includes

npx langfuse-cli api: discover the schema, then list, get and change resources
Start with `api __schema`, then `api <resource> --help`. Reads and writes go to your Langfuse project using your API keys.
Try: “List my last 20 traces with errors in Langfuse and summarize the failures.”
Runs code · Langfuse API via CLI
llms.txt index, pages as markdown, and docs search
Claude is told never to implement from memory. It fetches current pages from langfuse.com, or the API reference served by your self-hosted deployment.
Network · Documentation lookup
Add tracing to an existing function or app
references/instrumentation.md, always using the latest SDK version unless you say otherwise.
Write · Instrumentation
Create datasets, set up evals, calibrate an LLM judge, analyze errors
Guides: create-dataset, setting-up-evals, judge-calibration and error-analysis.
Write · Datasets and evals
Migrate prompts into Langfuse and edit them with engineering practice
Guides: prompt-migration and prompt-engineering.
Write · Prompts
Capture feedback as scores and cut LLM cost
Guides: user-feedback (ratings, behavior events, outcomes) and cost-optimization.
Write · User feedback and cost
langfuse/experiment-action checks and v4 project preparation
The CI/CD guide targets GitHub Actions and confirms evaluators and thresholds with you first. A separate guide prepares projects for the v4 platform.
Write · CI/CD gates and v4 migration

From the official docs

Permissions

Medium risk

This skill uses your Langfuse API keys to read and change project data, and runs the langfuse-cli npm package through npx. Use a key for a test project first and approve every change.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

It can change Langfuse data through the API
The CLI covers the full Langfuse REST API. The allowed-tools list pre-approves schema, help, list and get calls only, so other actions should ask first. Use project keys with the access you want to give. github.com
It runs a package from npm each time
The CLI is run with npx langfuse-cli, which downloads and runs that package. Pin a version in your own scripts if you need repeatable runs. github.com
Your search terms go to langfuse.com
Doc search sends the query to langfuse.com/api/search-docs. Keep customer data and secrets out of queries. github.com
Skill feedback is opt-in
When you say the skill is wrong, Claude offers to send feedback. The guide says to show every field and get your explicit permission first. It can use the Langfuse API or MCP, or a prefilled GitHub discussion link for you to submit. github.com

Technical details

Talks to	langfuse.com (docs, llms.txt, search). Your Langfuse server (cloud.langfuse.com, us.cloud.langfuse.com or your self-hosted URL)
Sign-in and keys	Keys come from environment variables or a .env file that you set; SKILL.md says not to paste keys into chat.
Scopes	Runs npx langfuse-cli (or bunx) with your LANGFUSE_PUBLIC_KEY and LANGFUSE_SECRET_KEY from the environment Fetches langfuse.com docs and its docs search with WebFetch and curl Edits your application code when you ask for instrumentation

Questions

What does the langfuse skill do?

It makes Claude work with Langfuse: instrument an app for tracing, migrate prompts, build datasets and evaluators, query traces and scores, and look up current docs before writing code.

How do I install it?

Use the install box at the top of this page. The Langfuse README also documents npx skills add langfuse/skills --skill "langfuse", a Cursor plugin and a manual symlink.

Do I need a Langfuse account and API keys?

Yes. You need a Langfuse project on cloud or self-hosted, plus a public and secret key from Settings > API Keys. The README says to export them as environment variables. A free cloud account can be created at langfuse.com/cloud, per SKILL.md.

Is it safe to use?

It uses your API keys to call your Langfuse project and runs langfuse-cli through npx. Start with a test project, keep keys out of chat and approve changes one by one.

Does it work in Codex or Cursor?

The README describes it for Claude Code, Cursor and similar assistants, and offers a Cursor plugin. It does not list other clients.

What is the difference from migrate-to-langfuse?

The README describes migrate-to-langfuse as a separate skill for moving from LangSmith, Arize, Braintrust and others, including a live tracing cutover. The langfuse skill covers day-to-day tracing, prompts, evals and queries.

Information

updated Nov 2, 2026

Repository: langfuse/skills
Folder: skills/langfuse

langfuse/skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 2

github.com · github.com
github.com · github.com