MLflow

What is MLflow?

MLflow is a long-running open-source MLOps platform that now covers LLM and agent apps too. It records traces of LLM calls and agent steps (OpenTelemetry-compatible), runs evaluations with LLM judges, versions prompts in a prompt registry, and includes an AI gateway for routing requests to model providers. The classic features remain: experiment tracking and a model registry. A single tracking server with a web UI serves the whole team; it runs on SQLite by default and scales to PostgreSQL with artifacts in S3, GCS or Azure Blob Storage.

Where it fits in a 10–200 person business

• Trace what a support chatbot or agent did on each request and where it went wrong
• Compare prompt or model changes with repeatable evaluations before rolling them out
• Track fine-tuning or classic ML experiments and keep a registry of approved model versions

These use cases are our editorial suggestions, not claims made by the project.

Can a business use MLflow? (license)

This is a plain-English summary of the license text, not legal advice. Third-party models you run with it have their own licenses.

Data & compliance: does MLflow phone home?

Checked against the project's own docs and repository on Nov 1, 2026. "Not documented" means we could not find the answer in official sources; it does not mean the feature is missing or that no data is sent. Third-party services you connect (model APIs, vector databases, search engines) have their own data policies. Compare all apps.

What hardware does MLflow need?

The project does not publish a minimum CPU/RAM figure, so we leave the official value blank rather than guess.

GPU: No.

No CPU, RAM or disk requirements are published. The quickest setup is a single process with SQLite and local artifact storage; the Docker Compose bundle adds PostgreSQL and RustFS (S3-compatible storage).

Source: mlflow.org/docs/latest/self-hosting/

How much does it cost to host MLflow per month?

How to deploy MLflow

Method: Docker Compose (MLflow + PostgreSQL + RustFS)

git clone https://github.com/mlflow/mlflow.git
cd mlflow/docker-compose
cp .env.dev.example .env
# change the default passwords in .env first
docker compose up -d
# open http://localhost:5000
# quickest alternative: pip install mlflow && mlflow server --port 5000

From the official instructions: mlflow.org/docs/latest/self-hosting/. Always check the source for the current version before you run it.

Known gotchas when self-hosting MLflow

• Authentication is off unless you start the server with --app-name basic-auth. That mode needs MLFLOW_FLASK_SERVER_SECRET_KEY and an admin password in MLFLOW_AUTH_ADMIN_PASSWORD set before the first start, or the server refuses to start. source
• Anonymized usage data is collected by default since version 3.2.0. Set MLFLOW_DISABLE_TELEMETRY=true or DO_NOT_TRACK=true to opt out. source
• The Compose example .env ships default credentials (POSTGRES_PASSWORD=mlflow, AWS_SECRET_ACCESS_KEY=s3admin) and binds MLflow to 0.0.0.0. Change them before exposing the server. source
• The built-in protection against DNS rebinding, CORS attacks and clickjacking (3.5.0+) only works with the default uvicorn server, not with --gunicorn-opts or --waitress-opts. When serving on 0.0.0.0, set --allowed-hosts to your real hostnames. source

Alternatives in LLM gateway & observability

• LiteLLM AI gateway that puts 100+ LLM providers behind one OpenAI-compatible API with keys and budgets.
• New API Self-hosted AI API gateway (One API fork) that puts many model providers behind one key, with quotas, usage logs and cost tracking.
• One API Original self-hosted LLM API gateway that puts many model providers behind one OpenAI-style key, with quotas and token management.
• Langfuse LLM observability with tracing, prompt management, evaluations and cost tracking.
• Opik Open-source LLM observability and evaluation platform for tracing, evals, prompt management and monitoring.
• Phoenix Free self-hosted LLM tracing, evaluation, datasets and prompt management from Arize.

Sources

• GitHub repository: github.com/mlflow/mlflow
• Official documentation, Hardware requirements, Deploy instructions: mlflow.org/docs/latest/self-hosting/
• License text: github.com/mlflow/mlflow/blob/master/LICENSE.txt
• Gotcha 1: mlflow.org/docs/latest/self-hosting/security/basic-http-auth/
• Gotcha 2: mlflow.org/docs/latest/community/usage-tracking/
• Gotcha 3: github.com/mlflow/mlflow/tree/master/docker-compose
• Gotcha 4: mlflow.org/docs/latest/self-hosting/security/network/
• Hetzner Cloud pricing (cost estimate): www.hetzner.com/cloud/regular-performance
• GitHub stars, commits and releases: GitHub REST API, fetched Nov 1, 2026