Docs Getting started
Quick start
Sign in, upload a dataset, ask your AI to do something with it.
ST-1 is an analytics workbench you and your AI share. Both of you operate on the same DuckDB-backed surface — uploads, profiles, SQL, and charts — so the AI’s work is something you can read, check, and edit instead of vouch for.
1. Sign in
Authenticate with GitHub from the landing page. Each project is private to you and shareable by URL.
2. Create a project
From the projects index, click New project. A project is a workspace — sources, derived tables, and visualizations all live inside it. Lineage stays inside the project boundary.
3. Upload data
Inside a project, click + Upload in the sidebar. CSV, TSV, Parquet, and JSONL are all accepted; ST-1 ingests with DuckDB and stores the result as Parquet. The schema is profiled automatically: nulls, distinct counts, ranges, hygiene flags.
4. Connect your AI
Open Connected agents and copy the MCP endpoint into Claude Desktop, Cursor, or any client that speaks Model Context Protocol. Your AI now sees every project you own as a tool surface. See Connect your AI over MCP for the full walk-through and the list of tools your AI can call.
5. Ask
From your AI: “Profile this dataset and create a chart of the most common categories.” The AI runs the upload, profiles columns, and creates a visualization in your project. Each action shows up in your browser as it happens.