MCP Server

The Model Context Protocol (MCP) Server exposes your site’s capabilities as structured tools that Claude and other MCP-compatible AI agents can call directly. This is the highest-fidelity integration layer — 99% reliable, 10–100x faster than visual automation.

What is MCP?

MCP (Model Context Protocol) is an open standard for connecting AI models to external tools and data sources. Instead of browsing your website visually, an AI agent calls your tools directly with structured input and receives structured output.

Traditional (Visual CUA):
  Claude → Opens browser → Navigates to site → Clicks buttons → Reads text → Responds

MCP (Direct Tool Calling):
  Claude → Calls get_product(sku: "ABC-123") → Receives JSON → Responds

Why MCP?

Available Tools

When you enable the MCP Server, the following tools are automatically generated from your site’s capabilities:

get_product

Retrieve detailed product information including pricing, inventory, and variants.

{
  "name": "get_product",
  "description": "Get product details by ID or SKU",
  "input_schema": {
    "type": "object",
    "properties": {
      "id": { "type": "string", "description": "Product ID or SKU" }
    },
    "required": ["id"]
  }
}

search_products

Search across the product catalog with filters.

{
  "name": "search_products",
  "description": "Search products by query, category, or price range",
  "input_schema": {
    "type": "object",
    "properties": {
      "query": { "type": "string" },
      "category": { "type": "string" },
      "minPrice": { "type": "number" },
      "maxPrice": { "type": "number" },
      "limit": { "type": "number", "default": 10 }
    },
    "required": ["query"]
  }
}

get_inventory

Check real-time inventory for a product or variant.

submit_lead

Submit a lead form with structured validation.

check_availability

Check appointment or booking availability.

Setup

1. Install the MCP SDK

npm install @inception-agents/mcp

2. Create your MCP server

import { createMcpServer, buildToolsFromManifest } from '@inception-agents/mcp';

const server = createMcpServer({
  name: 'your-store',
  version: '1.0.0',
  apiKey: process.env.INCEPTION_API_KEY,
});

// Auto-generate tools from your Express Flows manifest
const tools = await buildToolsFromManifest({
  manifestUrl: 'https://your-domain.com/agent/manifest',
});

server.registerTools(tools);
server.start();

3. Configure Claude Desktop

Add your MCP server to Claude Desktop’s configuration:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "your-store": {
      "command": "npx",
      "args": ["@inception-agents/mcp", "--api-key", "iak_your_api_key"],
      "env": {
        "INCEPTION_API_URL": "https://inception-agents-api.inception-agents.workers.dev"
      }
    }
  }
}

4. Auto-Discovery

The MCP server supports auto-discovery. When Claude encounters your domain, it can automatically detect and connect to your MCP server via the /.well-known/mcp.json endpoint:

{
  "name": "your-store",
  "version": "1.0.0",
  "description": "Product catalog and ordering tools for Your Store",
  "endpoint": "https://your-domain.com/mcp",
  "tools": ["get_product", "search_products", "get_inventory", "submit_lead"]
}

Tool Schema Generation

Tools are automatically generated from your:

1. Commerce connector data — Product catalog, inventory, collections
2. Express Flows — Any registered flow becomes an MCP tool
3. Custom definitions — Define additional tools programmatically

import { createMcpServer } from '@inception-agents/mcp';

const server = createMcpServer({ name: 'my-store' });

// Custom tool
server.tool('compare_products', {
  description: 'Compare two products side by side',
  input: {
    productA: { type: 'string', required: true },
    productB: { type: 'string', required: true },
  },
  handler: async ({ productA, productB }) => {
    const [a, b] = await Promise.all([
      getProduct(productA),
      getProduct(productB),
    ]);
    return { comparison: buildComparison(a, b) };
  },
});

Verification & Billing

All MCP tool calls are:

• Authenticated via API key
• Cryptographically signed for tamper-proof billing
• Logged in the dashboard under Analytics → MCP Interactions
• Billed only for verified, successful interactions

Security

• Tool calls require a valid API key
• Rate limiting applies per-key (see Rate Limits)
• Sensitive operations (checkout, payments) require additional verification
• All tool schemas are validated against input before execution

For the full SDK reference, see the @inception-agents/mcp README.