🚀6K+ orders
🇮🇳No hidden charges
₹500 Credit

Public MCP server for AI agents

Public, read-only Model Context Protocol (MCP) endpoint for crowcrowcrow.com. Lets Claude, ChatGPT, Cursor, Perplexity and custom agents search products, browse categories, and read blog/CMS content via structured tool calls — no API key, no scraping.

What you can do with it

  • Search and recommend CrowCrowCrow products inside a chat assistant.
  • Compare two products and surface ratings, prices, and key features.
  • Read blog posts and CMS pages (about, shipping, returns…) without scraping.
  • Build a shopping agent that links the user back to the PDP for checkout.

Tools

products.searchproducts

Keyword search; returns compact summaries (use products.get for full detail).

products.listproducts

Browse the catalogue with category, brand, price, and country filters.

products.getproducts

Full product detail by slug or sku — description, images, FAQs, related.

products.relatedproducts

Related published products for a given slug (curated + category fallback).

categories.listcategories

Six top-level categories with live product counts and PLP URLs.

categories.treecategories

Browse-path tree (1–4 levels deep) with counts.

categories.brandscategories

Top brands overall or scoped to a category.

blog.listblog

Published blog posts (newest first), paginated.

blog.getblog

Single published post by slug — sanitized HTML body.

pages.listpages

All published static CMS pages (about, shipping, returns, etc.).

pages.getpages

Single CMS page by slug — sanitized HTML body.

All tools are documented in the live WebMCP descriptor and the OpenAPI spec. Capable MCP clients also auto-load three resource documents (ccc://guide/readme, ccc://schema/product, ccc://guide/categories) so the agent has the schema before it calls a tool.

Limits (anti-scraping)

The MCP exists to make our catalogue discoverable, not harvestable. Every anonymous request is bounded:

LimitValue
Items returned per call (limit)20 (default 10)
Reachable pagination (offset)200
Anonymous requests per minute / IP60
Long text fieldsTruncated to 4 000 characters
Long arrays (highlights, FAQs…)Capped at 12 items

When a rate limit is hit you get an HTTP 429 with a Retry-After header and a JSON-RPC error message. Need higher caps for a partner integration? Email support@crowcrowcrow.com for a Bearer token.

Connect from Claude Desktop / Cursor

Add the snippet below to your claude_desktop_config.json (Claude Desktop) or your Cursor MCP settings. It uses the official mcp-remote bridge so HTTP MCP servers appear as if they were local.

{
  "mcpServers": {
    "crowcrowcrow": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://crowcrowcrow.com/api/mcp/mcp"]
    }
  }
}

Connect from Claude.ai (web) / ChatGPT custom connectors

Both clients accept a remote MCP URL directly. Paste https://crowcrowcrow.com/api/mcp/mcp into their Add Connector / Add MCP server dialog. Auth: None.

Try it from curl

1. Initialize the session

curl -s -X POST https://crowcrowcrow.com/api/mcp/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize",
       "params":{"protocolVersion":"2024-11-05","capabilities":{},
                  "clientInfo":{"name":"my-app","version":"1.0"}}}'

2. Search products

curl -s -X POST https://crowcrowcrow.com/api/mcp/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call",
       "params":{"name":"products.search",
                  "arguments":{"query":"protein bar","limit":5}}}'

3. Fetch full product detail

curl -s -X POST https://crowcrowcrow.com/api/mcp/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":3,"method":"tools/call",
       "params":{"name":"products.get",
                  "arguments":{"slug":"earth-mama-vegan-nipple-butter-2-pack"}}}'

Response shape (summary)

products.search and products.list return compact summaries:

{
  "slug": "earth-mama-vegan-nipple-butter-2-pack",
  "sku": "AMZ-...",
  "asin": "B0...",
  "url": "https://crowcrowcrow.com/p/<slug>/<sku>",
  "title": "Earth Mama Vegan Nipple Butter (2-Pack)",
  "brand": "Earth Mama",
  "shortDescription": "...",
  "category": "Health & Supplement",
  "categoryPath": "Baby Products › Feeding › Nursing & Feeding",
  "price": { "currency": "INR", "price": 3587, "listPrice": 4601, "discountPercent": 22 },
  "rating": 4.8,
  "reviewCount": 529,
  "inStock": true,
  "primaryImage": "https://...",
  "countryStore": "USA"
}

products.get adds description, images[], highlights[], keyFeatures[], whyChoose[], howToUse[], productDetails[], faqs[], dimensions, country of origin, comparisonTable[], relatedSlugs[], comparableSlugs[]. All prices are in INR and already include applicable Indian import duties.

FAQ

Do I need an API key?

No. The MCP is anonymous. Just rate-limited.

Can I use this to scrape your full catalogue?

Not really — the offset cap of 200 per query limits how deep you can paginate, and the rate limit will cut you off long before you finish. The MCP exists for agentic discovery, not bulk export. For bulk feeds use Google Merchant XML or contact us for partner access.

Are the prices accurate?

Yes. The MCP runs the same USD→INR conversion the storefront UI uses (live rate from currency_rates), so values match what a human visitor sees on the PDP.

Are there write operations?

No. The public MCP is strictly read-only — there is no way to place orders, modify the catalogue, or sign up users through it. Checkout happens on the storefront with browser session auth + Razorpay.

How do I report a bug or request a tool?

Email support@crowcrowcrow.com with the request payload, the response, and what you expected. We publish changes via /.well-known/webmcp.json.

More agent resources