Each page of the search is one API call, and each call is also a typed MCP tool, so an AI agent can run the export. Send the search link as url and the page you want:
curl -X POST "https://app.gtm-api.com/linkedin/v4/api/linkedin-scraping/search-sales-nav-people" \
-H "Authorization: Bearer gtm_live_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"url": "https://www.linkedin.com/sales/search/people?query=...",
"page": 1, "linkedin_account_sid": "ln_ac_YOUR_ACCOUNT"}'
- Paging. Rows sit at
result.rows, 25 a page. Call again with the next page whileresult.paging.has_moreis true; it turns false on the last page, and page 100 is the ceiling.result.paging.totalis LinkedIn's own count of the search: exact on a small search, an estimate on a large one. - The row. The lead fields of the export above, plus
pending_invitation(the account that ran the search has an invitation out to the lead). - MCP. The tool is
scrape_linkedin_search_sales_nav_peopleonhttps://mcp.gtm-api.com/mcp. For a collector that pages to the end on its own and waits out the budget, an agent callscreate_linkedin_auto_scrapewith the link and reads the leads back withsearch_linkedin_auto_scrape_results. - Budget. A call over the account's daily budget answers
429withretry_after, the moment it can run again. An account without a seat answers422 sales_nav_requiredbefore anything runs.
The docs: People and company search, Collect leads with an auto-scrape, and the reference for the search call, creating an auto-scrape and reading its results. Connecting a client is on the MCP config generator.