AI & MCP Integration

Coolpie.ai documentation is designed to be natively accessible to AI assistants and LLM-powered developer tools — no scraping, no workarounds.

No authentication required. All endpoints described on this page are publicly accessible. The documentation itself is public, so the AI integration is too.

MCP Server

The Coolpie documentation MCP (Model Context Protocol) server lets AI assistants — primarily Claude — query the docs directly as part of a conversation. Instead of copy-pasting documentation into a chat, developers working on Coolpie integrations can have the AI look up the relevant sections automatically.

Endpoint

https://mcp.docs.coolpie.com

Property	Value
Protocol	MCP over HTTP (Streamable HTTP transport)
Authentication	None — public endpoint
Content	Full text of all documentation pages, updated on every docs deploy
Infrastructure	Cloudflare Workers — edge-deployed, low latency globally

How it works

The server bundles a pre-built index of all documentation pages (extracted from HTML at deploy time). When a tool is called, it searches or retrieves from this in-memory index — no database, no external calls, sub-100ms response times.

Available Tools

The MCP server exposes three tools. All tools are called via POST https://mcp.docs.coolpie.com/call with a JSON body {"tool": "...", "input": {...}}.

list_coolpie_pages

Returns a list of all available documentation pages with their slug, title, description, URL, and section (user-guide or technical).

Input: none

get_coolpie_page

Returns the full text content of a specific documentation page.

Input: slug (string, required) — page identifier, e.g. data-outputs, matching, price-suggestions

search_coolpie_docs

Full-text search across all documentation pages. Returns ranked results with a snippet showing the matching context and the URL of the relevant page.

Input: query (string, required) — natural language or keyword search query

Example: search call

// Request
POST https://mcp.docs.coolpie.com/call
Content-Type: application/json

{
  "tool": "search_coolpie_docs",
  "input": { "query": "json export format" }
}

// Response
{
  "result": [
    {
      "slug": "data-outputs",
      "title": "Data Outputs",
      "url": "https://docs.coolpie.com/data-outputs.html",
      "snippet": "...Coolpie.ai exposes its competitive pricing data via a JSON export file and Google BigQuery. Each website has a dedicated JSON file regenerated on every price monitoring cycle...",
      "score": 26
    }
  ]
}

Example: get_coolpie_page call

// Request
POST https://mcp.docs.coolpie.com/call
Content-Type: application/json

{
  "tool": "get_coolpie_page",
  "input": { "slug": "matching" }
}

// Response
{
  "result": {
    "slug": "matching",
    "title": "Matching",
    "url": "https://docs.coolpie.com/matching.html",
    "section": "user-guide",
    "content": "... full page text ..."
  }
}

Setup in Claude

Add the MCP server once in your Claude settings and it will be available in every conversation automatically.

Claude.ai (web)

Go to Settings → Integrations → Add MCP Server
Enter URL: https://mcp.docs.coolpie.com
Name it Coolpie Docs and save

Claude Desktop / Claude Code

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "coolpie-docs": {
      "url": "https://mcp.docs.coolpie.com"
    }
  }
}

Once connected, Claude can answer questions like "How does the JSON export work?" or "What fields are in a product entry?" by querying the docs directly — without you needing to paste anything into the chat.

llms.txt

For AI systems that use web crawling rather than MCP (ChatGPT, Gemini, Perplexity, etc.), the documentation index is available as a machine-readable llms.txt file at:

https://docs.coolpie.com/llms.txt

This follows the llms.txt standard — a plain-text index that tells LLM crawlers what pages exist, what they contain, and how they relate to each other. It is updated automatically on every documentation deploy.