Model Context Protocol

Make Your Website
AI-Agent Ready

Expose a standard MCP endpoint on your domain. Claude, ChatGPT, Perplexity, and every MCP-compatible agent gets structured access to your content and products — through 5 typed tools, not HTML scraping.

Start Free TrialRead the Docs

Three steps. Zero code changes.

Your origin server never sees MCP traffic.

01

Enable MCP

Toggle the MCP Server switch in your domain settings. The gateway starts serving /mcp, /.well-known/mcp.json, and /llms.txt.

02

Gateway Proxies

The edge proxy intercepts /mcp requests, authenticates via API key, and routes to the AI service with your domain context.

03

Agents Connect

AI agents discover your MCP endpoint automatically and start querying your content through 5 structured tools.

5 structured tools. One endpoint.

Every tool enforces multi-tenant isolation. Queries are scoped to your domain automatically.

search_content

Content Search

Semantic search across every crawled page — blog posts, policies, FAQs, docs. Returns ranked results with title, URL, and similarity score.

get_page

Page Retrieval

Get the full markdown content of any page by URL path. Clean, structured text with metadata — ready for AI reasoning.

search_products

Product Search

Natural language product catalog search. Returns title, price, availability, brand, category, and relevance score.

get_product_details

Product Details

Full enriched product profile with 30+ AI-analyzed attributes — specs, FAQ, highlights, materials, dimensions, and more.

list_categories

Category Browser

Browse the complete product category hierarchy. Lets agents understand catalog structure before searching.

Auto-discoverable by every AI agent.

When MCP is enabled, the gateway automatically serves two discovery endpoints. AI agents find your MCP server without any manual configuration.

/.well-known/mcp.json

JSON manifest with server info, endpoint URL, authentication requirements, and available tool descriptions. The standard discovery path for MCP clients.

/llms.txt

Plain text discovery file following the emerging llms.txt standard. Describes your site and points agents to the MCP endpoint for structured access.

Connect in seconds.

Add your site to Claude Desktop with three lines of config.

claude_desktop_config.json
{
  "mcpServers": {
    "your-site": {
      "url": "https://your-domain.com/mcp",
      "headers": {
        "Authorization": "Bearer sw_prod_..."
      }
    }
  }
}

What MCP replaces.

Without MCP
With SerpWise MCP
Manual HTML scraping
Structured semantic search
Custom API development
Standard MCP protocol
Outdated product data
Real-time enriched catalog
No AI agent access
Auto-discoverable endpoint
Per-site integration work
One toggle to enable
Unstructured page content
Clean markdown with metadata

Decision support

Questions & Answers

What is MCP and why does my site need it?

The Model Context Protocol (MCP) is an open standard from Anthropic that lets AI agents connect to data sources via structured tools. Without MCP, AI agents must scrape your HTML — losing structure, missing products, and hallucinating details. With MCP, they get direct access to your content through typed, searchable tools. Your site becomes a first-class data source for every AI assistant.

Which AI agents can connect to my MCP server?

Any MCP-compatible client — Claude Desktop, ChatGPT (via plugins/tools), Perplexity, Cursor, Windsurf, custom agents built with the MCP SDK, and more. The protocol is open and growing. Discovery endpoints (/.well-known/mcp.json and /llms.txt) help agents find your server automatically.

Do I need to change my origin code?

No. The MCP endpoint is served entirely by the SerpWise edge proxy. Your origin server never sees MCP traffic. Enable MCP in your domain settings and the gateway handles authentication, tool routing, and response formatting.

How does authentication work?

AI agents authenticate with a Bearer token — an API key you create in the SerpWise dashboard with the products:read scope. Keys can be scoped to specific domains and have configurable rate limits. The gateway validates every request before routing to the AI service.

What data can AI agents access?

Agents can search all crawled pages (blog posts, policies, FAQs, product pages), retrieve full page content as markdown, search and browse the product catalog with 30+ AI-enriched attributes, and list product categories. All queries are scoped to your organization — no cross-tenant data leakage.

Does MCP work with non-ecommerce sites?

Yes. The search_content and get_page tools work with any crawled content — SaaS docs, agency portfolios, news sites, anything. Product-specific tools (search_products, get_product_details, list_categories) activate when product intelligence data is available.

Make your site AI-agent ready today.

One toggle. Five structured tools. Every AI agent can query your content and products in seconds.

Start Free TrialRead the Docs