AI Potluck
Product / UX / Telemetry & observability

MLflow

Databricks

AI engineering platform covering the machine-learning and LLM lifecycle in one tool, combining distributed tracing of agents and model calls, automated evaluation, a prompt registry and an AI gateway with the experiment tracking and model registry it began as. Databricks Managed MLflow is sold beside it as hosting rather than as a paid tier.

The product entry does not declare its GitHub repository, so its adoption band routes through PyPI only. Verified 2026-08-13 via the mlflow/mlflow LICENSE file, the repository and mlflow.org.

Openness

5 high confidence
5.0
license
Apache-2.0
source
public
core-gated
ungated

Apache-2.0 across a fully published core; Databricks Managed MLflow is third-party hosting rather than a gated core.

  • https://github.com/mlflow/mlflow/blob/master/LICENSE.txt recorded 2026-08-13

    The LICENSE file body, read rather than assumed. It carries the Databricks copyright line followed by the Apache License Version 2.0 text in full, including the "How to apply the Apache License to your work" appendix.

  • https://github.com/mlflow/mlflow recorded 2026-08-13

    Repository page for the published platform. 27,501 stars, source public, one root license.

  • https://mlflow.org/ recorded 2026-08-13

    Project site, describing "The leading open source AI engineering platform" with Observability, Evaluations, Prompt Registry, AI Gateway and Model Training as its features. Nothing is named as withheld for a paid tier.

Adoption

5 high confidence
5.0

MLflow draws 42,050,608 PyPI downloads in the last 30 days, well clear of the >10M floor that sets level 5 on the download scale - for comparison, pydantic-ai at roughly 31M and langgraph at roughly 58M also sit at 5. It is the de facto open experiment-tracking and ML-lifecycle platform, and the figure clears the floor comfortably.

Capability

4 high confidence
4.0

A single open platform combining distributed tracing, prompt versioning, automated evaluation and trace replay for agents and LLMs on top of classic ML experiment tracking. There is no standard benchmark for observability, so the score is set on the feature matrix, and MLflow sits level with Langfuse - the two of them are the pair the rest of this category is placed against.

  • https://mlflow.org/genai/ recorded 2026-08-13

    "Built-in observability, evaluation, prompt management, and monitoring. 100+ integrations", with the feature nav reading Observability, Evaluations, Prompt Registry, AI Gateway and Model Training, and the LLMs & Agents copy offering "production-grade tracing, evaluation, prompt management, and much more".

  • https://mlflow.org/top-5-agent-observability-tools/ recorded 2026-08-13

    "Top 5 LLM and Agent Observability Tools in 2026", the vendor's own comparison piece placing MLflow among the category leaders. Vendor-authored and read as positioning rather than as an independent ranking.

Verified 2026-08-13