ColdIQ MCP Server

An MCP (Model Context Protocol) server that exposes ColdIQ's B2B data marketplace to AI assistants. Connect it to Claude or any MCP-compatible client to search companies, find and enrich contacts, verify emails, track sales signals, and more — all powered by ColdIQ's unified API.

Prerequisites

• Node.js 18+
• A ColdIQ API key

Configuration

| Variable | Required | Default | Description | |---|---|---|---| | COLDIQ_API_KEY | Yes | — | Your ColdIQ API key | | COLDIQ_API_URL | No | https://api.coldiq.com | Override the API base URL | | COLDIQ_DEBUG | No | — | Set to 1 to enable debug logging |

Connecting to Claude

Claude Desktop

Add the following to your claude_desktop_config.json:

{
  "mcpServers": {
    "coldiq": {
      "command": "npx",
      "args": ["-y", "@coldiq/mcp@latest"],
      "env": {
        "COLDIQ_API_KEY": "your_api_key_here"
      }
    }
  }
}

Claude Code

claude mcp add coldiq --scope user --trust -e COLDIQ_API_KEY=<your-key> -- npx -y @coldiq/mcp@latest

The --scope user flag makes the server available in all your projects. The --trust flag skips per-tool approval prompts.

Running directly

COLDIQ_API_KEY=your_key npx @coldiq/mcp@latest

Available Tools

Prospecting

| Tool | Description | |---|---| | search_companies | Search B2B companies by firmographics, tech stack, funding, headcount, or hiring signals | | find_people | Find people at companies filtered by job title or department | | find_influencers | Discover LinkedIn or Twitter influencers in a niche |

Contact Data

| Tool | Description | |---|---| | find_email | Find a professional email from a name and domain (waterfall across providers) | | verify_email | Check if an email address is deliverable | | find_phone | Find a phone number for a contact |

Enrichment

| Tool | Description | |---|---| | enrich_person | Enrich a person's profile from a LinkedIn URL, name, or domain | | enrich_company | Enrich a company's profile from its domain, LinkedIn URL, or name |

Intelligence & Research

| Tool | Description | |---|---| | find_signals | Surface sales signals: funding rounds, acquisitions, job changes, hiring intent, buying intent, news | | search_jobs | Search live job postings; routes to ATS career sites or LinkedIn based on filters | | search_ads | Discover active LinkedIn ads from a company | | search_seo | SERP rankings, traffic data, and technology stack analysis | | search_reddit | Search Reddit posts and discussions |

Web & Local

| Tool | Description | |---|---| | search_web | General web search (Google or neural/Exa) | | search_places | Find local businesses and places | | fetch_page_content | Extract text content or a summary from any URL |

Development

# Run unit tests
npm test

# Watch mode
npm run test:watch

# Live integration tests (consumes credits)
LIVE_TESTS=1 npm run test:gtm

Tests live in tests/, mirroring the src/ structure. See tests/gtm-scenarios.md for realistic end-to-end usage scenarios.

Architecture

The server is built on three layers:

• Tools (src/tools/) — MCP tool definitions with Zod input schemas and descriptions
• Executor (src/executor.ts) — Waterfall/fallback engine that tries providers in priority order
• Registry (src/registry.ts) — Maps each capability to an ordered list of providers with parameter translation, applicability gates, and response normalization