Leadpipe MCP

Official MCP server for Leadpipe intent and data APIs.

Use this server to give agents direct access to:

• topic discovery

• website topic analysis

• audience preview and ad hoc query

• saved audience management

• audience status, results, runs, stats, and exports

• visitor data lookup

• tracking pixel management

• account health and credit status

What This Is

Leadpipe MCP wraps the public Leadpipe intent and data APIs in a local MCP server that agents can use over stdio.

This server is API-key authenticated by default so the full audience workflow, visitor data access, and pixel management workflows are available to agents.

Related MCP server: MessengerFlow MCP Server

Use With Claude, Cursor, and Codex

Most users do not need to clone this repo or build from source. The published package is:

npx -y @leadpipe/mcp

You will need:

export LEADPIPE_API_KEY=sk_...

Optional advanced override:

export LEADPIPE_BASE_URL=https://api.aws53.cloud

LEADPIPE_BASE_URL only needs to change if you are pointing at a non-default environment.

Claude Desktop

Add this to your Claude Desktop MCP config:

{
  "mcpServers": {
    "leadpipe": {
      "command": "npx",
      "args": ["-y", "@leadpipe/mcp"],
      "env": {
        "LEADPIPE_API_KEY": "sk_..."
      }
    }
  }
}

Cursor

Add this to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "leadpipe": {
      "command": "npx",
      "args": ["-y", "@leadpipe/mcp"],
      "env": {
        "LEADPIPE_API_KEY": "sk_..."
      }
    }
  }
}

Codex

Add the server with the Codex CLI:

codex mcp add leadpipe --env LEADPIPE_API_KEY=sk_... -- npx -y @leadpipe/mcp

Then confirm it is configured:

codex mcp list

Quick Start

If you just want to verify the package runs:

npx -y @leadpipe/mcp

Run From Source

For local development or repo hacking:

npm install
npm run build
node dist/index.js

For iterative local development:

npm run dev

Local source build MCP config:

{
  "mcpServers": {
    "leadpipe": {
      "command": "node",
      "args": ["/absolute/path/to/mcp/dist/index.js"],
      "env": {
        "LEADPIPE_API_KEY": "sk_..."
      }
    }
  }
}

Tool Surface

Topic Discovery

• list_topics

• get_topic_facets

• search_topics

• get_topic_trend

• compare_topics

• get_topic_movers

• analyze_website_topics

Audience Builder

• get_audience_filters

• preview_audience

• query_audience

• list_audiences

• get_audience

• create_audience

• update_audience

• delete_audience

Audience Results

• get_audience_status

• get_audience_results

• list_audience_runs

• get_audience_stats

• export_audience

Visitor Data

• query_visitor_data

Pixels

• list_pixels

• create_pixel

• update_pixel

Account

• get_account_status

Prompts

• discover-audience-topics

• operate-saved-audience

• investigate-visitor-data

Prompt Examples

See examples/prompt-examples.md for copy-paste prompts that work well with this MCP server.

Good starting prompts:

• "Find 5-10 B2B topics for cloud cost optimization software, compare their recent trends, and preview an audience."

• "Create a saved audience for these topics, activate it, wait until it is ready, and then summarize the first page of results."

• "Check the status of audience <audience-id>. If it is ready, export it and give me the download URL."

• "Investigate visitor activity for example.com over the last 30 days and tell me whether our tracking setup looks healthy."

• "List our pixels, find any paused ones, and explain what needs attention."

Agent Playbooks

For more structured operational flows, see:

• Audience Research Playbook

• Visitor Diagnostics Playbook

Resources

• leadpipe://docs/auth

• leadpipe://docs/workflows

• leadpipe://docs/data-api

Typical Agent Workflow

1. Search or analyze topics

2. Preview audience size

3. Create or update a saved audience

4. Check status until the audience is ready

5. Fetch results or export CSV

Typical Data Workflow

1. Check get_account_status

2. Query visitor activity with query_visitor_data

3. Inspect or create pixels with list_pixels / create_pixel

4. Pause or reactivate tracking with update_pixel

Development

npm run typecheck
npm run build

License

MIT