id, title, body, source, and source-specific metadata such as the original URL or filing date, so results are easy to store, cite, and pass to a model.
Ask from Claude
Add Dataville to Claude as a connector, then ask in plain language:edgar source and answers with the form type, the filing date, and headline figures from the filing. The same connector works in Claude Code, Cursor, Windsurf, and VS Code. See Quickstart: Claude for the setup.
Try it from the browser
No account or API key is needed. Pick a source, type keywords, and press send:Example responsepress Send for live data
data is always the single most relevant result. Try edgar with a company name for its latest SEC filing and headline financials, or arxiv with a paper title.
Beyond single lookups
The same sources are available in other ways.- Claude, Cursor, Windsurf, and VS Code can search every source over MCP. See Connect an AI tool.
- You can download a whole source as CSV, Parquet, or a DuckDB database. See List available exports. Parquet exports support range requests, so DuckDB on your own machine can read them directly without downloading the file.
- You can query across sources with SQL. Run a SQL SELECT across the datasets runs a read-only
SELECTover every exportable source, with joins across sources, up to 1,000 rows per query.
Next steps
Quickstart: Claude
Add the connector and ask for a company’s most recent filing.
Quickstart: API
Create an API key and make your first request from code.
Connect an AI tool
Set up Dataville in Claude Code, Cursor, Windsurf, or VS Code.
Demos
Try live apps built on Dataville and get the code for each one on GitHub.