MLflow
Open-source tracing, evaluation and prompt management for GenAI and ML.
The long-standing ML lifecycle platform now covers GenAI: OpenTelemetry-compatible tracing for agents, LLM-judge evaluations, prompt versioning and a model registry, self-hosted or managed on Databricks.
- Vendor
- Databricks (Linux Foundation project)
- Category
- Observability & evals
- Pricing
- Open source
- Open source
- Yes
- License
- Apache 2.0
- Platforms
- Python, TypeScript, Self-host, Cloud
- Launched
- 2018
- Website
- mlflow.org
Features
- Agent tracing
- LLM evaluation
- Prompt registry
- Model registry
Best for
- Enterprise
More observability & evals
- LangSmith — LangChain. Tracing, evals and monitoring for LLM apps.
- Langfuse — Langfuse. Open-source LLM engineering platform.
- Braintrust — Braintrust. The evals platform for AI products.
- promptfoo — promptfoo. Test and red-team your LLM apps.
- Arize Phoenix — Arize AI. Open-source AI observability and evaluation.
- W&B Weave — Weights & Biases. Track and evaluate LLM applications.
- Helicone — Helicone. LLM observability via a one-line proxy.
- Ragas — Exploding Gradients. Evaluation metrics for RAG pipelines and agents.