MCP Server

The official DataBlue Model Context Protocol server gives MCP-compatible AI clients a scraper-first DataBlue toolset: search, scrape, batch scrape, extraction, crawling, and docs-ready Data APIs. Version 1.1 runs locally over stdio and authenticates with your DataBlue API key.

Run the server

npx --yes --prefer-online @datablue/mcp@latest

The @latest and --prefer-online flags check npm when the MCP process starts. Restart the client to pick up a new published version.

Claude, Cursor, and Cline

{
  "mcpServers": {
    "datablue": {
      "command": "npx",
      "args": ["--yes", "--prefer-online", "@datablue/mcp@latest"],
      "env": {
        "DATABLUE_API_KEY": "<your-key>"
      }
    }
  }
}

Codex

Add this block to your Codex config.toml:

[mcp_servers.datablue]
command = "npx"
args = ["--yes", "--prefer-online", "@datablue/mcp@latest"]

[mcp_servers.datablue.env]
DATABLUE_API_KEY = "<your-key>"

Core tools

Tool	Purpose
datablue_search_google	Search Google and return structured results.
datablue_scrape_url	Scrape one URL into markdown, HTML, links, structured output, or screenshots through formats: ["screenshot"].
datablue_batch_scrape	Scrape up to 20 selected URLs with the same options and return one full result row per URL.
datablue_extract	Extract schema-shaped data from URLs or supplied content.
datablue_map_site	Discover URLs available on a site.
datablue_start_crawl	Start an asynchronous site crawl.
datablue_get_job_status	Check a crawl or other asynchronous job.
datablue_list_data_apis	List stable, docs-ready Data APIs and their parameters.
datablue_run_data_api	Run a listed Data API by catalog id.

Resources

Clients can inspect read-only context before choosing a tool: datablue://apis, datablue://pricing, datablue://usage, datablue://docs/scrape, datablue://docs/crawl, datablue://docs/google-serp, datablue://examples/batch-scrape, and datablue://examples/research-company.

Workflow prompts

The server includes research_company, competitor_serp_report, extract_products_from_url, find_leads_from_google, scrape_site_to_markdown, and monitor_keyword.