# MLflow

> MLflow is the leading open-source AI engineering platform for debugging, evaluating, monitoring, and optimizing LLM applications, AI agents, and ML models.

MLflow is the largest open-source AI engineering platform, trusted by thousands of organizations with over 30 million monthly downloads. It covers the full lifecycle of AI development — from LLM and agent observability to classical ML experiment tracking — under a single Apache 2.0-licensed platform. Built on OpenTelemetry and supporting 100+ integrations, MLflow works with any cloud, framework, or LLM provider without vendor lock-in.

- **Observability & Tracing**: *Capture complete traces of LLM applications and agents using OpenTelemetry-compatible instrumentation; monitor production quality, costs, and safety in real time.*
- **LLM Evaluation**: *Run systematic evaluations with 50+ built-in metrics and LLM judges, track quality over time, and catch regressions before they reach production.*
- **Automatic Issue Detection**: *Use AI-powered analysis to automatically detect issues across correctness, latency, execution, adherence, relevance, and safety dimensions in your traces.*
- **Prompt Registry & Optimization**: *Version, test, and deploy prompts with full lineage tracking; automatically optimize prompts using state-of-the-art algorithms.*
- **AI Gateway**: *Unified OpenAI-compatible API gateway for all LLM providers — route requests, manage rate limits, handle fallbacks, and control costs.*
- **Agent Server**: *Deploy agents to production with a single command using a FastAPI-based hosting solution with streaming support and built-in tracing.*
- **Experiment Tracking**: *Log parameters, metrics, and artifacts for ML experiments; compare runs and reproduce results with ease.*
- **Model Registry & Deployment**: *Manage the full ML model lifecycle from staging to production with a centralized model registry and deployment tools.*
- **Broad Framework Support**: *Integrates natively with OpenAI, Anthropic, LangChain, LlamaIndex, CrewAI, AutoGen, PyTorch, HuggingFace, and 100+ more frameworks.*
- **Multi-language Support**: *Supports Python, TypeScript/JavaScript, Java, and R, making it accessible across diverse engineering teams.*

To get started, install MLflow via pip, launch the tracking server with `uvx mlflow server`, add a single `mlflow.openai.autolog()` call to your code, and explore traces and metrics in the MLflow UI.

## Features
- LLM & agent observability with OpenTelemetry-compatible tracing
- LLM evaluation with 50+ built-in metrics and LLM judges
- Automatic issue detection across correctness, latency, safety, and relevance
- Prompt Registry with versioning and lineage tracking
- Prompt optimization with state-of-the-art algorithms
- AI Gateway with unified OpenAI-compatible API
- Agent Server for one-command production deployment
- ML experiment tracking with parameter and metric logging
- Model Registry and deployment tools
- 100+ integrations with AI frameworks and LLM providers
- Multi-language support: Python, TypeScript/JavaScript, Java, R
- Production monitoring for quality, costs, and safety
- Human feedback collection for LLM applications
- Apache 2.0 open-source license

## Integrations
OpenAI, Anthropic, LangChain, LangGraph, LlamaIndex, Vercel AI, Amazon Bedrock, LiteLLM, Gemini, Google ADK, Strands Agent, DSPy, PydanticAI, Agno, Semantic Kernel, AutoGen, CrewAI, PyTorch, HuggingFace, OpenTelemetry, Databricks

## Platforms
LINUX, WEB, API, DEVELOPER_SDK, CLI

## Pricing
Open Source

## Links
- Website: https://mlflow.org
- Documentation: https://mlflow.org/docs/latest/
- Repository: https://github.com/mlflow/mlflow
- EveryDev.ai: https://www.everydev.ai/tools/mlflow