Get started

Four short guides take you from nothing to a trained model you can explain and use. Each one stands on its own and takes 5 to 15 minutes.

	Guide	Time	You'll end with
1	Run ParleML	5 min	ParleML open in your browser, with an agent signed in
2	Your first model	10 min	A churn model, its accuracy on unseen rows, and what drives it
3	Work with an agent	15 min	An agent that frames, builds and explains a model while you decide
4	A tour of ParleML	5 min	The handful of ideas every page of the app is built on

Then, when you need them:

• Use your model in Python: install the scoring package and predict in a few lines of code.
• Use ParleML from Claude Code: let Claude Code explore your data and train models through ParleML's MCP server.

No data of your own yet?

You don't need any. ParleML ships eight sample datasets, one for each kind of problem: churn, fraud with text notes, house prices, sales forecasting, customer personas, unusual payments, product baskets and next-best product. Every guide here uses them.

Where to go after this

• Tutorials: longer walkthroughs of one problem from start to finish.
• How-to guides: one job each, such as connecting a database or scoring a file.
• User manual: every page of the app, explained.