Connect an Agent
The Commerce for Agents MCP server is reachable at https://mcp.commerceforagents.com/mcp over Streamable HTTP. There is no authentication and no API key: add the endpoint and call the tools. This page has short setup snippets for common clients and SDKs. Full tool documentation is at llms.txt.
Claude Desktop / claude.ai (custom connector)
In Settings → Connectors → Add custom connector, paste the endpoint URL. No API key field is needed.
https://mcp.commerceforagents.com/mcp
Claude Code
Add the server with the CLI; it uses Streamable HTTP.
claude mcp add --transport http commerce-for-agents https://mcp.commerceforagents.com/mcp
Cursor
Add an entry to ~/.cursor/mcp.json (or Settings → MCP).
{
"mcpServers": {
"commerce-for-agents": {
"url": "https://mcp.commerceforagents.com/mcp"
}
}
}
VS Code
Add an entry to .vscode/mcp.json. VS Code uses a servers key and an explicit type.
{
"servers": {
"commerce-for-agents": {
"type": "http",
"url": "https://mcp.commerceforagents.com/mcp"
}
}
}
OpenAI Agents SDK (Python)
Use MCPServerStreamableHttp as an async context manager and pass it to the agent's mcp_servers.
from agents import Agent, Runner
from agents.mcp import MCPServerStreamableHttp
async def main():
async with MCPServerStreamableHttp(
params={"url": "https://mcp.commerceforagents.com/mcp"},
) as server:
agent = Agent(
name="assistant",
instructions="Use the connected tools to help the user.",
mcp_servers=[server],
)
result = await Runner.run(agent, "Which home internet providers serve 350 5th Ave, New York, NY 10118?")
print(result.final_output)
LangChain (langchain-mcp-adapters)
MultiServerMCPClient takes a per-server config; use transport: "streamable_http".
from langchain_mcp_adapters.client import MultiServerMCPClient
client = MultiServerMCPClient({
"commerce-for-agents": {
"url": "https://mcp.commerceforagents.com/mcp",
"transport": "streamable_http",
}
})
tools = await client.get_tools()
LlamaIndex (llama-index-tools-mcp)
BasicMCPClient wraps the endpoint; McpToolSpec turns it into LlamaIndex tools.
from llama_index.tools.mcp import BasicMCPClient, McpToolSpec
client = BasicMCPClient("https://mcp.commerceforagents.com/mcp")
tools = await McpToolSpec(client=client).to_tool_list_async()
Raw JSON-RPC (curl)
The endpoint speaks plain MCP over HTTP: a session starts with initialize, then tools/list works without further handshake state on this server.
curl -s https://mcp.commerceforagents.com/mcp \
-H 'Content-Type: application/json' \
-H 'Accept: application/json, text/event-stream' \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
Returns a tools array with all seven tools, each with its full JSON Schema input: compare-tariffs, tariff-details, start-booking, compare-mobile-plans, book-contract, submit-switch-request and compare-broadband. The two booking tools are for the booking form and for autonomous agents that already hold the user's data and authorization.
Notes
- Transport is Streamable HTTP; SSE-only clients should look for a Streamable HTTP or "HTTP" transport option, not "SSE".
- No authentication, no API key, no OAuth. See auth.md.
- Tool descriptions are in English; the tools themselves answer in the language of the conversation.
- Machine-readable references: llms.txt, MCP server card, OpenAPI document.