>MLflow - Open Source AI Platform for Agents, LLMs & Models

# MLflow > MLflow is the largest open source AI engineering platform for agents, LLMs, and models. MLflow enables teams of all sizes to debug, evaluate, monitor, and optimize their AI applications while controlling costs and managing access to models and data. With over 30 million monthly downloads, thousands of organizations rely on MLflow each day to ship AI to production with confidence. MLflow provides two main feature sets: 1. **LLMs & AI Agents**: Production-grade observability (tracing), evaluation with LLM judges, prompt management, an AI Gateway for managing costs and model access, human feedback collection, and agent serving. 2. **Machine Learning**: Experiment tracking, hyperparameter tuning, model evaluation, a model registry for version control and deployment management, unified model packaging, and flexible model serving. Supports classical machine learning and deep learning models. MLflow supports all LLM providers, AI agent frameworks, and coding assistants, including LangChain, LangGraph, OpenAI, Anthropic, ADK, Pydantic AI, Claude Code, Codex, Cursor, DSPy, CrewAI, Gemini, DeepSeek, GLM, Kimi, and beyond. It works on any major cloud provider (AWS, Azure, GCP, Databricks) or on-premises infrastructure. Native SDKs are available for Python, TypeScript/JavaScript, Java, and R, and MLflow's REST API integrates seamlessly with any programming language. MLflow is backed by the Linux Foundation and is 100% open source under the Apache 2.0 license. ## Website - [MLflow Home](https://mlflow.org/): Overview of the MLflow AI engineering platform - [LLMs & Agents](https://mlflow.org/genai): Ship AI agents and LLM apps to production with built-in observability, evaluation, prompt management, and monitoring - [Observability](https://mlflow.org/genai/observability): End-to-end observability for agents and LLM applications with execution visualization and tracing - [Evaluations](https://mlflow.org/genai/evaluations): Assess agent and LLM output quality with