MCP Server

QuantSearch exposes an MCP server, so an AI agent can create sites, ingest and remove content, run searches and read analytics without you writing an API client. It speaks JSON-RPC over HTTP POST.

Endpoint

Environment	URL
Production	`https://mcp.quantsearch.ai`
Staging	`https://mcp.stage.quantsearch.ai`

Authentication

Same API keys as the REST API — Authorization: Bearer qs_…. Every tool exceptsignup requires one. See Authentication for how to create a key.

Claude Desktop / Claude Code

{
  "mcpServers": {
    "quantsearch": {
      "url": "https://mcp.quantsearch.ai",
      "headers": {
        "Authorization": "Bearer qs_your_api_key"
      }
    }
  }
}

Raw JSON-RPC

curl -X POST https://mcp.quantsearch.ai \
  -H "Authorization: Bearer qs_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

Tools

Getting started

Tool	What it does
`signup`	Create an organization and return an API key. The one tool that needs no auth, for agents starting cold. Rate limited to 3/hour per IP.
`whoami`	Confirm which organization a key belongs to

Sites

Tool	What it does
`list_sites`	List sites in the organization
`create_site`	Create a site and start its first crawl
`get_site_status`	Crawl and index status for one site
`update_site`	Change settings, public access, rate limits
`recrawl_site`	Re-crawl, optionally with one-shot overrides
`delete_site`	Delete a site and its index
`get_embed_snippet`	Widget or search-page embed code

Content

Tool	What it does
`upsert_pages`	Push pages straight into the index, bypassing the crawler. Idempotent by URL. For Astro/Hugo/CMS sites where the agent already holds the content.
`delete_pages`	Remove pages by exact `urls` and/or glob/regex `patterns`. At least one is required.
`list_pages`	List what is indexed; verify an upsert or delete landed
`search`	Run a search against a site

Removing a whole section without listing every URL:

{
  "site_id": "4226a896-f2d3-40d3-aa80-5bc627d4c83b",
  "patterns": ["/blog/archive/*"]
}

Account

Tool	What it does
`get_analytics`	Search volume and trends
`get_top_queries`	Most frequent queries for a site
`get_usage`	Usage against plan limits
`start_checkout`	Begin a plan upgrade
`open_billing_portal`	Link to the billing portal

Ingested pages are not auto-removed

Stale-content reconciliation only covers pages the crawler found. A page pushed in withupsert_pages stays indexed until you remove it, even if it later disappears from the source site — use delete_pages when your CMS deletes content.