Web Research — tool reference

Server URL: https://toolcargo.com/mcp/web-research · Scope: connector:web-research.

Activate Site Audit free, then connect Firecrawl. Your key is checked against the credit balance endpoint, stored encrypted and deleted on disconnect. ToolCargo only sends authenticated requests to api.firecrawl.dev.

Every successful tool call uses one shared Site Audit call. Firecrawl provider credits are separate, paid through your Firecrawl account. Failed ToolCargo calls are refunded; a timeout may still consume provider credits. No automatic retries are performed.

web_search

Titles, URLs and snippets for a query. Up to 10 results, default 5. Search does not automatically scrape the results.

{ "query": "Saudi ecommerce market report", "limit": 5 }

web_scrape

One public HTTP(S) page as clean Markdown, with title, description and up to 50 links. Text defaults to 12,000 characters; max 30,000. No private/local URLs, credentials, login cookies, custom request headers, browser actions or PDF parsing options. The provider handles rendering and redirects; this is not the Site Audit fetcher.

{ "url": "https://example.com", "max_chars": 12000 }

web_map

Find public page URLs on a website. Default 25, max 100; an optional search term filters discovery. A URL map is not a full content crawl. Results identify when the requested limit was reached.

{ "url": "https://example.com", "search": "pricing", "limit": 25 }

firecrawl_credit_usage

Remaining/plan credits and provider billing period. No scraping is performed; the tool uses one ToolCargo call.

All tools return JSON text. Page content is untrusted external material; it should be used as source content, not instructions. Requests have a 35-second timeout and a 4 MB response cap. No results are saved by ToolCargo. Firecrawl processes the query/URL under its own service policies.

Implementation references: official Firecrawl MCP and Firecrawl API documentation. ToolCargo uses an original API adapter compatible with its MCP SDK v2.