Skip to content
Back to all tools

MLflow

Open-source AI engineering platform for agents, LLMs and ML models

Consider with caveatsAPIFree tier
Jack Phillips
Audited by Jack Phillips · Updated May 2026
Visit site

Overall score

3.6/ 5
SME fit4/5
flat pricing + free tier · technical setup
JTBD4/5
solid named JTBD
Integration4/5
API + 10 integrations
Trust5/5
mature, founded 2018
Quality1/5
no public rating
Compliance2/5
customer-choice residency

About

MLflow is the open-source platform for managing the AI/ML lifecycle, originally released by Databricks in 2018 and now used across more than 60 million monthly downloads. The current product spans classical ML (experiment tracking, model registry, deployment) and modern LLM and agent workflows (tracing, evaluation, prompt versioning). Self-host the framework for free under Apache 2.0, or use Managed MLflow billed through Databricks.

Best for: Engineering teams shipping LLM agents or production ML models who want open-source experiment tracking, evaluation and observability without paying for a closed MLOps SaaS.

Pricing

  • Open Source

    Monthly
    Free
    Annual /mo
    Free
    Billing
    flat
    Notes
    Full framework feature set;Apache 2.0 license;Self-hosted tracking server;Model registry;LLM tracing and evaluation;All integrations · Install via pip install mlflow; you operate the tracking backend and artifact store.
  • Managed MLflow (via Databricks)

    Monthly
    n/a
    Annual /mo
    n/a
    Billing
    usage_based
    Notes
    Fully managed tracking server;Model registry with Unity Catalog governance;LLM tracing and evaluation;Production agent monitoring;14-day Databricks trial with $400 in credits · Billed via per-second Databricks DBU consumption; rates vary by cloud provider and workload type.

Key features

  • Experiment tracking with parameters, metrics and artifacts
  • Model registry with versioning and stage transitions
  • LLM tracing and observability built on OpenTelemetry
  • 50+ built-in LLM evaluation metrics plus LLM-as-a-judge
  • Prompt versioning and optimization
  • Model and agent deployment as REST APIs or batch inference

Integrations

OpenAIAnthropicLangChainLlamaIndexDSPyHugging FacePyTorchTensorFlowscikit-learnDatabricks

Trust & compliance

Stage range
MVP → Growth
Founded
2018
Status
active
SOC 2
unknown
GDPR
unknown
Data residency
customer_choice
External rating
n/a
Last verified
May 2026

Reviews

Be the first to share your experience.

Pairs well with