AskRobots Search

AskRobots Search MCP server

Web search from an independent index, for AI agents. AskRobots Search has its own crawler and index; results come from that crawl, not a live fetch and not from another engine.

The server is hosted. This repository holds its listing for the MCP Registry (server.json) and these setup notes. The server's source is not public.

• Endpoint: https://search.askrobots.com/api/v1/mcp
• Transport: Streamable HTTP. JSON answers only (no SSE stream), stateless.
• Auth: an API key, sent as Authorization: Bearer <key> or X-API-Key: <key>.
• Docs: https://search.askrobots.com/api, and the REST API's OpenAPI file at https://search.askrobots.com/api/v1/openapi.json

Pricing

From https://search.askrobots.com/api:

• $25 a month, including 500 searches. After that, $0.05 a search.
• We email you the key, usually within a day. It is shown once; keep it out of code you publish.

The keyless trial (10 searches a day per address) is for the REST endpoint /api/search only. The MCP server always needs a key; without one it answers 401.

Tools

Seven tools, all read-only.

Tool	What it does
search	Web results with title, URL, snippet and why each matched, plus related searches; ten a page, with site, freshness, news, language and verbatim options, and optionally the best passages of the top results.
answer	Our stored AI summary of a search, with numbered sources to cite; the top results when there is none yet.
page	What we know about a URL, or up to 25: title, description, language, page type, when we fetched it, health, and how many sites link to it.
site	What we know about a website: topic, directory listing, pages indexed, linking sites, and flags (parked, blocklisted, adult, junk, in a link network).
links	The sites that link to a page or a site, most trusted first, with a link network counted once; up to 50 a call.
read	Our stored copy of a page: with a query, the passages that answer it; without, the outline and text in parts. Pages that refuse copies are refused.
browse_directory	Our directory of sites by topic, ranked by links from other listed sites.

Safe search is always on.

Set up

Put your key in an environment variable, here ASKROBOTS_API_KEY, rather than in a file you commit.

Claude Code

claude mcp add --transport http askrobots-search https://search.askrobots.com/api/v1/mcp \
  --header "Authorization: Bearer $ASKROBOTS_API_KEY"

Claude Desktop

Claude Desktop's config file starts local servers, so connect through mcp-remote (needs Node.js). In claude_desktop_config.json:

{
  "mcpServers": {
    "askrobots-search": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://search.askrobots.com/api/v1/mcp",
        "--header",
        "X-API-Key:${ASKROBOTS_API_KEY}"
      ],
      "env": {
        "ASKROBOTS_API_KEY": "YOUR_ASKROBOTS_API_KEY"
      }
    }
  }
}

Cursor

In ~/.cursor/mcp.json (or .cursor/mcp.json in a project):

{
  "mcpServers": {
    "askrobots-search": {
      "url": "https://search.askrobots.com/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer ${env:ASKROBOTS_API_KEY}"
      }
    }
  }
}

VS Code

In .vscode/mcp.json. VS Code asks for the key once and stores it securely:

{
  "inputs": [
    {
      "type": "promptString",
      "id": "askrobots-api-key",
      "description": "AskRobots API key",
      "password": true
    }
  ],
  "servers": {
    "askrobots-search": {
      "type": "http",
      "url": "https://search.askrobots.com/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer ${input:askrobots-api-key}"
      }
    }
  }
}

Other clients

Any client that speaks Streamable HTTP can connect: give it the endpoint and one of the two headers above.

Contact

[email protected] · https://search.askrobots.com/about

License

MIT, see LICENSE. The license covers this repository's files; the hosted service has its own terms: https://search.askrobots.com/terms