Open-source MLOps platform for experiment tracking and model deployment.
## MLflow MCP Server: Complete ML Lifecycle Platform The **MLflow MCP Server** integrates the open-source MLflow platform into Google Antigravity. This comprehensive tool manages the entire ML lifecycle from experimentation to production deployment, ensuring reproducibility and collaboration. ### Why MLflow MCP? MLflow standardizes ML workflows: - **Experiment Tracking**: Log parameters, metrics, artifacts - **Model Registry**: Version and stage models - **Model Serving**: Deploy models to production - **MLflow Projects**: Reproducible ML code - **Open Source**: Works with any ML library ### Key Features #### 1. Experiment Tracking ```python import mlflow # Start experiment mlflow.set_experiment("classification-experiments") with mlflow.start_run(): # Log parameters mlflow.log_param("learning_rate", 0.01) mlflow.log_param("epochs", 100) # Train model model = train_model(lr=0.01, epochs=100) # Log metrics mlflow.log_metric("accuracy", 0.95) mlflow.log_metric("f1_score", 0.94) # Log model mlflow.sklearn.log_model(model, "model") ``` #### 2. Model Registry ```python # Register model from run model_uri = f"runs:/{run_id}/model" mlflow.register_model(model_uri, "ProductionClassifier") # Transition model stage client = mlflow.tracking.MlflowClient() client.transition_model_version_stage( name="ProductionClassifier", version=1, stage="Production" ) # Load production model model = mlflow.pyfunc.load_model( "models:/ProductionClassifier/Production" ) ``` #### 3. Model Serving ```python # Serve model as REST API # mlflow models serve -m models:/MyModel/Production -p 5000 # Or deploy to cloud mlflow.deployments.create_deployment( target="sagemaker", name="my-deployment", model_uri="models:/MyModel/Production" ) ``` ### Configuration ```json { "mcpServers": { "mlflow": { "command": "npx", "args": ["-y", "@anthropic/mcp-mlflow"], "env": { "MLFLOW_TRACKING_URI": "http://localhost:5000", "MLFLOW_S3_ENDPOINT_URL": "your-s3-endpoint" } } } } ``` ### Use Cases **Experiment Management**: Compare hundreds of experiments to find the best model configuration. **Model Versioning**: Track model lineage and easily rollback if production issues arise. **Team Collaboration**: Share experiments and models across data science teams. The MLflow MCP Server brings enterprise ML lifecycle management to Antigravity.
{
"mcpServers": {
"mlflow": {}
}
}