MCP Tools

Tools

Tools are schema-validated actions the agent can call. API keys are never accepted as tool input.

Tool	Required input	Limits and defaults
`datablue_search_google`	`query`	1-100 results, 1-5 pages; country `us`, language `en`, Lite mode by default
`datablue_scrape_url`	`url`	Markdown by default; supports HTML, links, screenshots, selectors, extraction, waits, and proxy routing
`datablue_batch_scrape`	`urls`	1-20 URLs with one shared scrape configuration
`datablue_extract`	`urls`, `prompt`	1-20 URLs; prompt up to 10,000 characters; optional JSON Schema, provider, and model

Tool	Required input	Limits and defaults
`datablue_map_site`	`url`	1-1000 links; optional search filter; subdomains and sitemap enabled by default
`datablue_start_crawl`	`url`	1-1000 pages, depth 1-10, concurrency 1-10; returns a job ID
`datablue_get_job_status`	`job_id`	`job_type` is `crawl`, `search`, or `data`
`datablue_get_result_items`	`result_id`	Read 1-100 cached rows with offset, optional object path, and optional field projection; zero DataBlue credits

Tool	Use
`datablue_list_data_apis`	Call without arguments for all stable API IDs and compact parameter schemas. Pass `api` for one API's full descriptions, example, limits, and live credit weight. Catalog reads use zero credits.
`datablue_run_data_api`	Run one API ID returned by the catalog with a `params` object.

Example

{ "api": "zepto_products" }

Unpublished or SDK-only IDs are rejected. Poll an asynchronous Data API with datablue_get_job_status and job_type: "data".

Example

Search Google for "DataBlue web scraping", scrape the first three organic results as Markdown, and extract each product name and source URL.