What the integration does
Download web files
Stagehand finds and downloads PDF and image files in a Browserbase session.
Store documents in Box
The workflow uploads each file to a folder owned by a Box service account.
Extract structured data
Box AI extracts fields, confidence scores, and source references. Structured extraction can apply OCR to images and scanned PDFs.
Route exceptions
A deterministic check compares the extracted fields and marks each file as
APPROVED or NEEDS_REVIEW.How it works
- Stagehand opens the manufacturer and government source pages in a Browserbase browser.
- The browser downloads the SDS and label PDFs to Browserbase.
- The workflow retrieves the original file bytes through the Browserbase Downloads API.
- Box stores both files and answers a cited safety question about the SDS.
- Box AI extracts structured fields from each file, including text embedded in the label artwork.
- The workflow compares the EPA registration numbers and checks the SDS revision date.
- Box metadata records the extracted values, Browserbase session ID, source URL, document role, and compliance status.
Box AI processes the SDS and label separately. The workflow compares their structured results in application code because Box AI doesn’t process a mixed text-and-image packet as one request.
What you’ll need
- A Browserbase API key
- A Box Server app that uses Client Credentials Grant
- A Box plan with access to the Box AI API
- A Box folder shared with the app’s service account
- Node.js 18 or newer and pnpm
Get started
Follow the quickstart
Configure Browserbase and Box, then run the compliance intake.
View the example
Review the complete Stagehand and Box implementation.