langfuse

To add LLM tracing, route instrumentation requests through this Skill to access reference-driven setup practices for capturing application execution data. It validates LANGFUSE credentials and guides tracing implementation to debug observability gaps in LLM behavior.

Migrate hardcoded prompts into versioned Langfuse prompts by routing read and write operations through the Langfuse MCP server. This enables fetching, creating, and labeling prompt versions via MCP tool calls within a unified management layer.

Yes, capturing user feedback as scores is supported by routing data access requests to the Langfuse CLI. You can list and inspect traces, observations, scores, datasets, and sessions to evaluate LLM application performance.

Wire Claude Code transcripts into Langfuse traces by configuring a Claude Code Stop-hook. This Skill provides targeted reference materials for hook setup to automatically capture and route transcript execution data into your observability backend.

Yes, Langfuse credential validation is required before routing prompt operations to the MCP server. The Skill checks LANGFUSE_* credentials to enable prompt reads and writes, falling back to the CLI for non-prompt resource access.