Connect an MCP client
CommerceAI speaks the Model Context Protocol over Streamable HTTP, so most MCP clients can connect with a URL and a header.
| Setting | Value |
|---|---|
| Server URL | https://commerceai.ai/mcp/ |
| Transport | Streamable HTTP |
| Auth header | Authorization: Bearer <your key> |
Client configuration
Claude Desktop uses the mcp-remote bridge to connect to the remote server. Add its configuration to claude_desktop_config.json, then restart Claude Desktop. For Cursor, use ~/.cursor/mcp.json or your project's .cursor/mcp.json.
{
"mcpServers": {
"commerce-ai": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://commerceai.ai/mcp/",
"--header",
"Authorization: Bearer ${COMMERCEAI_API_KEY}"
],
"env": {
"COMMERCEAI_API_KEY": "<YOUR_API_KEY>"
}
}
}
}{
"mcpServers": {
"commerce-ai": {
"url": "https://commerceai.ai/mcp/",
"headers": {
"Authorization": "Bearer <YOUR_API_KEY>"
}
}
}
}claude mcp add --transport http commerce-ai \
https://commerceai.ai/mcp/ \
--header "Authorization: Bearer <YOUR_API_KEY>"For the Python MCP SDK example and the complete response object, see Quickstart.
Verify the connection
Ask your agent something that needs product data:
Find me a stainless steel French press under $40 and compare the top three.
The client should show a shopping_agent tool call followed by a list of products.
Prompting tips
Include a budget, a category and one or two must-have attributes. Specific queries return tighter results and use fewer tokens in the agent's context.