MLflow
About this tool
An open-source ML and model-engineering platform for experiment parameters, metrics and artifacts, model versions, and LLM application tracing and evaluation. Since 3.5.1, an experimental MCP server lets agents search traces and record feedback. MLFLOW_MCP_TOOLS can enable experiment, run, model and deployment tools, allowing agents to query and compare experiment records.
- Feature 1: Track parameters, metrics and artifacts and compare runs.
- Feature 2: Manage model versions and aliases in the model registry.
- Feature 3: GenAI tracing and evaluation with scorers and human feedback.
- Feature 4: Built-in search_traces, get_trace and log_feedback tools, selectable through genai, ml or all tool groups.
- Feature 5: Use local storage, a self-hosted Tracking Server or Databricks hosting.
Resource information
- Maintainer
- MLflow, a Linux Foundation project led by Databricks.
- Type
- MCP
- Category
- Experiments & data analysis
- Runs on
- MCP
- Who it's for
- Groups training ML models, searching hyperparameters or developing LLM applications, using agents to compare runs, summarize results and review traces.
- License at time of listing
- Apache-2.0
- Source
- GitHub
