MLflow

What it does

MLflow is an open-source AI engineering platform for teams building or operating LLM applications, AI agents, and machine-learning models. Developers instrument application calls or send OpenTelemetry traces to an MLflow server; traces are grouped into experiments and can be inspected in the MLflow UI. The platform also supports systematic evaluations with built-in or custom metrics, prompt versioning and testing, and an AI Gateway for routing requests to model providers.

A typical tracing setup runs an MLflow server, creates an experiment, adds a client SDK and tracing instrumentation, then inspects application traces and metrics in the UI. The server can run locally or on infrastructure managed by the team. MLflow also provides experiment tracking, model evaluation, a model registry and agent-deployment tools. The Apache 2.0 software is free to use; hosting, compute and external model usage are separate operational costs. The Python quickstart requires Python 3.10 or later, and a Docker Compose option is available.

What it can do

• Capture and inspect traces from LLM applications and agents
• Evaluate applications with built-in or custom quality metrics
• Version, test, and deploy prompts
• Route model requests through an AI Gateway
• Track ML experiments and evaluate models
• Register models and deploy agents

Works with

• OpenTelemetry
• OpenAI
• LangChain
• PyTorch
• Anthropic
• Google
• Amazon Bedrock

Before you choose

• The quickstart's local Python setup requires Python 3.10+; the documented Docker option uses Docker Compose.
• Tracing requires an MLflow server and a client SDK or OpenTelemetry instrumentation; application traces are sent to that server.
• Self-hosting and external model providers may incur separate costs.

Listed under

• Coding
• Agents
• Developers
• Evaluation
• LLM
• Observability
• Open source

Get started

1. Install MLflow and start a server locally, for example with `pip install --upgrade mlflow` and `mlflow server`.
2. Create an experiment in the MLflow UI and install the client SDK for the application language and provider integration.
3. Set the tracking URI, enable tracing or instrument the application, run it, and inspect traces in the MLflow UI.

The project charges no software fee. Budget separately for hosting, compute and external model providers.

Source: mlflow.org, checked 1 Nov 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 1 Nov 2026. Each source supports part of the overview; availability and prices can change.

• MLflow - Open Source AI Platform for Agents, LLMs & Models ↗mlflow.org
• Tracing Quickstart | MLflow AI Platform ↗mlflow.org
• MLflow - Open Source AI Platform for Agents, LLMs & Models ↗mlflow.org

Company website

mlflow.org: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 1 Nov 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.