Try GroundX Studio | Documentation

Once GroundX Agent Harness is installed (see Install GroundX Agent Harness), see your agent put GroundX to work in about five minutes. The agent does the work: it creates the bucket and ingests the sample itself, so there are no manual dashboard steps.

Your agent needs to reach GroundX for these, through the hosted MCP tools or your API key. A chat-only agent can explain the steps but cannot run them.

1. Hand it a document

Have your agent set up a bucket and load a sample document, then tell you when it is ready to search. For example:

Set up a GroundX bucket called "my-first-bucket", add this invoice to it, and tell me when it's ready to search: https://example.com/sample-invoice.pdf

Point it at any document your agent can reach by URL, or a local path if your agent can read files.

2. Ask about it

What's the total due on that invoice, and which line does it come from?

3. Pull out the details

Pull the invoice number, date, vendor, each line item, and the total into a table.

4. Make extraction accurate at scale

For real volume, have your agent build an extraction workflow that grades itself against an answer key and tunes the schema and prompts until it is accurate:

Set up a bucket and an extraction workflow for these sample invoices: pull the invoice number, date, vendor, line items, and total from each. Then grade it against the answer key and keep tuning the schema and prompts until it's at least 90% accurate.

Go further

Full retrieval walkthrough (bucket, ingest, poll, search, answer): Ingest, Search, and Answer With an Agent.
Schema-first extraction in depth: Extract Data With an Agent.

Then point the agent at your own documents and go further: ask for a summary report, a way to classify them, or a small app to search them.