MCP Server Setup: Connecting Claude and GPT Agents to Cold Scout

The Model Context Protocol (MCP) is the open standard for letting AI agents call external tools. Cold Scout ships as an MCP server out of the box, so any MCP-aware client — Claude Desktop, Claude Code, Cursor, Cline, Continue — can ask the agent to discover, qualify, and draft cold emails directly.

What you can do with the integration

"Find 50 dental clinics in Mumbai with 4.5+ star ratings."
"Qualify those leads against this ICP and return the top 10."
"Draft a cold email to lead #42, referencing their about page."
"Run the daily pipeline and report what changed."

Quick setup (Pro plan)

Sign up for Cold Scout Pro — the MCP server is included.
Generate an API key from the dashboard's Settings → MCP page.
Add the server to your client's MCP config. For Claude Desktop, that's~/Library/Application Support/Claude/claude_desktop_config.json.
Restart the client. The Cold Scout tools appear in the agent's tool list.

Self-host setup (OSS)

The OSS package exposes the same MCP server athttp://localhost:8000/mcp by default. Point your client there with no auth for local-only setups, or terminate the public endpoint behind an authenticating reverse proxy for shared use.

Available tools

The server exposes the same primitives the dashboard uses internally:search_places, qualify_lead,generate_email, send_email, plus pipeline-level wrappers like run_discovery and run_qualification.

This guide is a stub

Full configuration examples, the JSON schema for each tool, and security recommendations for exposing the MCP server publicly are coming soon. In the meantime, the canonical reference is the /docs#mcp section of our technical documentation.

Quick setup (Pro plan)

Generate an API key from the dashboard's Settings → MCP page.

Add the server to your client's MCP config. For Claude Desktop, that's~/Library/Application Support/Claude/claude_desktop_config.json.

Restart the client. The Cold Scout tools appear in the agent's tool list.