Documentation
predictr.io can be driven from a few different surfaces — the docs mirror that. Pick whichever fits how you work; the underlying engine is the same.
Recommended for AI / agent users. If you plan to drive predictr through Claude Code (or any other agent), set up your data connections in the web app first and run an initial metadata scan there. After that, connections are referenced by name — Claude only ever sees the connection identifier, never your database credentials. Same applies to the CLI and direct API use.
Get going
- Walkthrough — your first analysis — a guided tour, start-to-finish in under ten minutes with a sample data file. The fastest way to get a feel for the product.
Use the web app
Step-by-step guides for the predictr.io web app.
- Creating an Analysis — overview of the data shapes each analysis needs.
- Market Basket Analysis — items that co-occur in orders or baskets.
- Customer Clustering — group customers by RFM-style behavioural attributes.
- Forecasting — time-series forecasts of revenue / demand / volume.
- Manage Existing Models — rebuild, retrain and inspect models you've already built.
Connect your data
- Creating Database Connections — overview of the connection types we support.
- Per-source reference: File Uploads, Snowflake, Google BigQuery, Amazon Redshift, Amazon Athena.
Other surfaces
- CLI — drive predictr from your shell.
- API — REST endpoints + auth.
- Claude Code — conversational access via the predictr plugin.
- Automated Workflows — schedule multi-step jobs (model rebuilds, scans, refreshes).