Sets up MLflow tracing for Python and TypeScript agents and LLM apps, with autoinstrumentation for LangChain, LangGraph, OpenAI, and other frameworks. The guide tells you what's actually worth tracing (LLM calls, retrieval, tool use) versus what adds noise (string formatting, config loading), which is more helpful than most observability docs. Includes verification steps to confirm traces are actually being logged before you waste time on evaluation, plus patterns for feedback collection and production deployment with sampling. Load this before running agent evaluation or you'll be debugging blind.
npx skills add https://github.com/mlflow/skills --skill instrumenting-with-mlflow-tracing