LUML Flow review
A local-first tracker with strong ML, LLM, evaluation, and registry coverage.
Reviewed by iTechGuides Editors · Editorial team · Updated Oct 2026
LUML Flow is a local-first experiment tracker for predictive machine learning and generative AI workflows. It records parameters, metrics, model artifacts, traces, evaluation samples, scorers, human annotations, and file attachments while experiments run. Its local UI supports cross-experiment comparison and stores experiment data in SQLite. Individual practitioners and ML teams can use the lumlflow Python package locally, then upload selected experiments to the LUML platform for collaboration, deployment, monitoring, and versioned model registry artifacts.
Its feature depth is strongest around experiment and evaluation workflows. Live parameter and metric tracking sits alongside model artifact logging, comparison across experiments, LLM trace capture through OpenTelemetry, evaluation samples, scorers, and annotations. The integration list covers PyTorch, XGBoost, scikit-learn, TensorFlow, LightGBM, JAX, Keras, LangChain, LangGraph, LlamaIndex, DSPy, and Anthropic. That breadth suits teams working across conventional ML and LLM applications, while local SQLite storage gives individual workflows a self-contained starting point.
The published plans show a free Light tier with all platform capabilities, 1 user, 1 orbit, 50 artifacts, 1 satellite, and community support. Plus costs €79 per month when billed annually and adds up to 3 users, 300 artifacts, 2 satellites, and standard support. Pro costs €249 per month when billed annually and raises those limits to 5 users, 1,000 artifacts, and 5 satellites. LUML Flow fits open-source teams that value local execution, broad framework coverage, and a path to a shared registry. Teams needing more than 5 users, 1 orbit, or 1,000 artifacts should compare alternatives before committing.
LUML Flow pros and cons
- Where it wins
- Tracks parameters, metrics, models, traces, evaluations, and attachments
- Supports comparison, SQLite storage, and a versioned model registry
- Integrates with PyTorch, LangChain, TensorFlow, and other ML tools
- Where it doesn't
- Light is limited to 1 user, 1 orbit, 50 artifacts, and 1 satellite
- Pro supports up to 5 users and 1,000 artifacts
- Platform access is centered on API and self-hosted deployment
LUML Flow fact sheet, pricing and score →
Advertiser disclosure: iTechGuides is reader-supported. Vendors can pay for top positions in our rankings and for a place on other products' pages, and we may earn a commission when you click some links. Bought places stand at the number bought; this page and our disclosure say so. Paying vendors may get a fuller, more detailed review from our editors, and that review can change their score. How we rank.
Last updated · How we research and update