OSS TanbouOverview
MLflow is an AI engineering platform covering machine-learning experiment tracking, model registry, evaluation, and deployment alongside agent and LLM tracing, evaluation, prompt management, monitoring, and AI Gateway capabilities. It provides a shared system for recording quality and lifecycle history across traditional ML and generative-AI applications.
Features and best fit
Based on official documentation; not hands-on tested · Content checked:
Track experiment parameters, metrics, artifacts, and models
Training runs can record parameters, metrics, artifacts, and models for comparison and lifecycle management through MLflow tracking and model-registry workflows.
Trace and evaluate agents and LLM applications
MLflow can capture application traces, run evaluations, manage and optimize prompts, and monitor production behavior for LLM and agent systems.
Sources: [2]
For teams unifying ML and generative-AI lifecycle records
MLflow fits organizations that want one platform to manage experiments, evaluations, traces, models, and prompts across classical ML and LLM/agent workloads.
Design production storage, authentication, and access explicitly
A local server is easy to start, but production deployments need deliberate backend and artifact storage, authentication, access controls, and backups. AI Gateway use also introduces provider credentials and cost governance.
Sources: [4]
Official sources
- [1]mlflow/mlflow repository(2026-09-20)
- [2]MLflow README(2026-09-20)
- [3]MLflow Apache-2.0 license(2026-09-20)
- [4]MLflow documentation(2026-09-20)
Supplemental curator note
MLflow has expanded well beyond traditional experiment tracking. Its current README covers agents and LLMs with tracing, evaluation, prompt management, monitoring, and an AI Gateway in addition to model training and registry workflows. Start with a local tracking server and evaluate the subset relevant to your AI lifecycle.
Try it in 3 steps
- 1
Install MLflow
Install MLflow in a virtual environment.
python -m pip install mlflow - 2
Start a local tracking server
Start the local server and open the UI at http://127.0.0.1:5000.
mlflow server --host 127.0.0.1 --port 5000 - 3
Log a parameter and metric
Create a minimal run and confirm parameters and metrics appear in the MLflow UI.
python -c "import mlflow; mlflow.set_tracking_uri('http://127.0.0.1:5000'); mlflow.set_experiment('demo'); run=mlflow.start_run(); mlflow.log_param('model','demo'); mlflow.log_metric('score',0.9); mlflow.end_run()"
Growth
Growth trends · Last 30 days
28,046 Stars
Trend data is still being collected.
Development activity
Last 90 days · weekly
- Commits (last 30 days)
- 416
- Open PRs
- 636
- Issues opened
- 336
- Issues closed
- 285
- PRs opened
- 1,503
- PRs merged
- 968
Issues
336 / 285
Pull requests
1,503 / 968
Maintenance
- Median first response
- 43 hr
- Issue response rate
- 56.7% (51/90)
Based on up to the 100 newest issues opened by external users in the last 90 days. A first comment from an OWNER, MEMBER, or COLLABORATOR counts as a response; issues whose full comment history cannot be checked are excluded. The median and response rate update weekly.
Built with
Categories and tags
Categories
GitHub data
GitHub dataView detailed GitHub data
GitHub Topics
- machine-learning
- ai
- ml
- mlflow
- apache-spark
- model-management
- agentops
- agents
- evaluation
- langchain
- llm-evaluation
- llmops
- Stars
- 28,046
- Forks
- 6,329
- Watchers
- 326
- Open issues
- 1,483
- Contributors
- 446
- Primary language
- Python
- License
- Apache-2.0
- Repository last updated
- Sep 20, 2026
Related information
Write a related articleShare a guide or use case for this OSS in Markdown. Articles are published after administrator approval.
Explore next
- Uptime Kuma91,570 Stars
A self-hosted monitoring tool for HTTP/TCP/DNS/Ping checks, notifications, and status pages.
JavaScript - Elasticsearch77,937 Stars
A distributed engine for full-text and vector search, analytics, logs, and metrics.
Java - Grafana76,812 Stars
An observability platform for visualizing metrics, logs, and traces from multiple sources and connecting them to alerts.
TypeScript - Prometheus66,130 Stars
A monitoring platform that pulls labeled time-series metrics and combines PromQL, rules, alerts, and service discovery.
Go - Sentry44,803 Stars
A debugging platform that connects error and performance detection to traces, runtime context, investigation, and fixes.
Python
Report incorrect information
Tell us if any listing information is incorrect or outdated.