ParleML

ParleML turns a business question into a model you can trust and explain. You can work by hand, page by page, or hand the work to an agent that frames the problem, builds and challenges models, and stops to ask you before anything that matters. Either way, every number links back to the evidence behind it.

New to ParleML and comparing platforms? Read Why ParleML first.

New here? Start with these

	Guide	Time
1	Run ParleML: one docker run, open your browser	5 min
2	Your first model: train and explain a churn model from a sample	10 min
3	Work with an agent: let an agent build a fraud model while you decide	15 min
4	A tour of ParleML: workspaces, datasets, experiments, models, sessions	5 min

Already running? Open Home, choose a sample under Start from a sample, and you have data to work with in one click.

Find your way

I want to…	Go to
Follow one problem from start to finish	Tutorials
Do one job: connect a database, score a file, set up sign-on	How-to guides
Use a trained model in my own Python code	Use your model in Python
Use ParleML from Claude Code, Cursor or Codex	Use ParleML from Claude Code
Understand a page of the app	User manual
Decide which kind of experiment fits my question	Which experiment fits?
Look up a setting, an environment variable or the API	Reference
Look up a term	Glossary
Fix a stuck session or a missing result	Troubleshooting

Go deeper

ParleML is built on three components, each with its own guide:

• ML4: training, feature engineering, templates and every configuration setting.
• MLI: data exploration, explanations, fairness and model cards.
• Agents: the MCP server, its tools, skills and sessions.

These guides are technical by design. You don't need them to use the app.