How MLflow provides coding assistants with the required feedback loop to build better LLM agents - trace, analyze, fix, validate, repeat.

mlflow

MLflow Skills is a new tool that gives coding agents (Claude Code, Codex, Gemini CLI, OpenCode) the context needed to build and improve LLM agents through a structured loop: set up tracing, run the agent, analyze traces for issues, write LLM-as-a-judge scorers to quantify problems, apply fixes, and verify improvements. The workflow eliminates manual evaluation by letting the coding agent handle trace analysis, scorer generation using MLflow's evaluation API, and regression checking. MLflow 3.10+ also supports multi-turn conversation evaluation. An MLflow Assistant in the UI provides the same loop without switching to the terminal.

Ship LLM Agents Faster with Coding Assistants and MLflow Skills

The Loop Inside the UI: MLflow Assistant ​