Skip to main content
At a high level, Bulkgrid separates direct retrieval from longer-running ingestion and extraction workflows. Search is the simplest path:
  1. your application sends a search request
  2. Bulkgrid returns ranked results immediately
  3. your application uses those results in UI, retrieval, or answer generation
This is the best first workflow to test because it does not require polling.

​
Asynchronous workflows: extract, crawl, and deep crawl

For extraction and crawl-style work, the flow is different:
  1. your backend submits a request
  2. Bulkgrid creates a run
  3. the run progresses through pending and processing
  4. results and content outputs are generated
  5. your backend polls for completion and retrieves outputs

​
Outputs customers consume

Depending on the workflow, customers consume:
  • ranked search results
  • structured extraction data
  • markdown output
  • clean HTML or raw HTML
  • links and metadata
  • screenshot access
The cleanest architecture is:
  • frontend calls your backend
  • your backend calls Bulkgrid with x-api-key
  • your backend stores run IDs and normalizes outputs for your own application model
This keeps secrets server-side and gives you control over retries, timeouts, and user-facing error handling.