Leadpipe MCP Server: Intent Data for AI Agents

Your AI agent can now talk directly to Leadpipe.

No API wrappers. No middleware. No custom integration code. One line:

npx -y @leadpipe/mcp

That gives Claude Desktop, Cursor, Codex, or any MCP-compatible agent access to 27 tools covering topic discovery, audience building, visitor data, pixel management, and account health.

We just shipped the Leadpipe MCP server. Here is what it does and how to set it up.

What is MCP?

MCP (Model Context Protocol) is the open standard for connecting AI agents to external tools and data. Instead of writing custom API integration code, you configure an MCP server and the agent gets structured access to every tool automatically.

Think of it as a USB port for AI. Plug in the Leadpipe MCP server, and your agent can search intent topics, build audiences, look up visitors, and manage pixels - all through natural language.

The protocol was created by Anthropic and is now supported by Claude Desktop, Cursor, Codex (OpenAI), and a growing number of agent frameworks.

What the Leadpipe MCP server exposes

27 tools across 6 categories. Your agent gets all of them:

Topic Discovery (7 tools)

Tool	What it does
`list_topics`	Browse the full catalog of 20,735 intent topics
`get_topic_facets`	Get filter options (industries, types, categories) with counts
`search_topics`	Search topics by keyword (autocomplete)
`get_topic_trend`	Daily trend data for any topic (audience size, scores, segments)
`compare_topics`	Side-by-side trends for up to 10 topics
`get_topic_movers`	Topics with biggest day-over-day audience growth or decline
`analyze_website_topics`	Paste any URL, get matched intent topics via AI extraction

Audience Builder (8 tools)

Tool	What it does
`get_audience_filters`	Available ICP filter values (seniority, industry, size, etc.)
`preview_audience`	Preview audience count + 50 masked samples before committing
`query_audience`	Run a full audience query with topics + ICP filters
`list_audiences`	List all saved audiences
`get_audience`	Get a single saved audience by ID
`create_audience`	Save a new audience definition
`update_audience`	Update name, config, or status (activate/pause)
`delete_audience`	Remove a saved audience

Audience Results (5 tools)

Tool	What it does
`get_audience_status`	Check if audience is materializing, ready, or failed
`get_audience_results`	Fetch full unmasked person profiles (paginated)
`list_audience_runs`	See daily run history with dates and counts
`get_audience_stats`	Field fill rates (e.g., email: 100%, LinkedIn: 83%)
`export_audience`	Generate CSV download URL (valid 24 hours)

Visitor Data (1 tool)

Tool	What it does
`query_visitor_data`	Look up identified visitors by email, domain, or timeframe

Pixel Management (3 tools)

Tool	What it does
`list_pixels`	List all tracking pixels and their status
`create_pixel`	Create a new pixel for a domain
`update_pixel`	Pause, activate, or set excluded paths

Account (1 tool)

Tool	What it does
`get_account_status`	Credits remaining, pixel count, audience slots, health check

Plus 3 built-in prompts that guide agents through common workflows:

discover-audience-topics - find topics, compare trends, preview an audience
operate-saved-audience - create, activate, wait, fetch results
investigate-visitor-data - check visitor activity, diagnose tracking setup

Setup: Claude Desktop

Add this to your Claude Desktop MCP config (claude_desktop_config.json):

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

Restart Claude Desktop. You will see Leadpipe tools available in the tools panel.

Setup: Cursor

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

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

Setup: Codex (OpenAI)

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

Confirm it is configured:

codex mcp list

Setup: Any MCP-compatible agent

The server runs over stdio. Point any MCP client at it:

LEADPIPE_API_KEY=sk_... npx -y @leadpipe/mcp

Prompt examples

Once the MCP server is running, you talk to your agent in natural language. It calls the right Leadpipe tools automatically. Here are real prompts you can copy and paste.

Topic discovery

Find strong B2B intent topics for procurement automation software.

Use Leadpipe topic discovery to:
1. search for relevant topics
2. compare recent trend lines
3. identify fast-growing movers
4. recommend the best 5 topics to use in an audience preview

Explain why each topic made the cut.

Analyze https://www.ramp.com and tell me which Leadpipe topics
best match the company.

Then suggest 3 adjacent topics I should also consider targeting.

Audience preview

Preview an audience for B2B topics related to cloud security
posture management.

Use these constraints:
- minimum score 70
- business email required
- LinkedIn required
- company size mid-market or enterprise

Return:
- estimated total audience size
- a short interpretation of the sample
- any warning if the audience looks too narrow

I want an audience for HR software buyers in California.

Find relevant topics, preview the audience, and tell me whether
the result set is broad enough for outbound.

Saved audiences

Create a saved audience named "Cloud Cost Buyers - EU" using the
best topics for cloud cost optimization.

After creating it:
1. activate it
2. check status until it is ready
3. summarize the first page of results

Do not export unless I ask.

Check audience <audience-id>.

If it is materializing, tell me the current status clearly.
If it is ready, show me:
- total count
- data date
- first few rows
- any obvious data quality observations

Exports

Export audience <audience-id>.

If a cached export already exists, reuse it.
Then give me the download URL and the row count.

For audience <audience-id>, list available runs first.
Then export the most recent ready run and tell me what date
it corresponds to.

Visitor data

Investigate visitor data for example.com over the last 30 days.

Use Leadpipe to:
1. check account health
2. query visitor activity for the domain
3. inspect configured pixels

Tell me whether tracking appears healthy and what I should fix
if not.

Look up visitor data for jane@company.com.

If there is a resolved journey, summarize:
- the domains involved
- recency
- anything notable about session activity

Pixel management

List all Leadpipe pixels for this account.

Group them into:
- active
- paused

Call out any domains that look misconfigured or duplicated.

Create a new pixel for www.example.com named "Example Marketing Site".

After creation, summarize:
- pixel id
- status
- domain

Pause pixel <pixel-id> and exclude these paths:
- /careers
- /blog

Then confirm the final pixel state.

Account health

Check Leadpipe account status and summarize:
- overall health
- credit usage
- pixel counts
- intent audience slot availability

Keep the summary concise and operational.

Tip: These prompts work best when the agent uses preview_audience before creating large audiences, checks get_audience_status before assuming results exist, and treats exports as delivery artifacts. More examples in the full prompt library on GitHub.

The typical agent workflow

Most agent sessions follow this pattern:

1. Search or analyze topics         (what are buyers researching?)
2. Preview audience size            (how many people match?)
3. Create or update saved audience  (save the query)
4. Check status until ready         (wait for materialization)
5. Fetch results or export CSV      (get the actual people)

The MCP server handles all of this. The agent never writes API code, parses JSON, or manages authentication. It just calls tools.

Why MCP instead of raw API?

You can already use the Leadpipe REST API or the TypeScript SDK (npm install @leadpipe/client) to build custom integrations.

MCP is different. It is not for your code. It is for your agent.

	Raw API / SDK	MCP Server
Who uses it	Your code	Your AI agent
Integration effort	Write code	Add 5 lines of config
Authentication	Handle in code	Configured once in env
Error handling	Your responsibility	Agent handles retries
New tools	Update your code	Update the package
Works with	Any language	Claude, Cursor, Codex, any MCP client

If you are building a product on Leadpipe data, use the SDK. If you want your AI agent to have Leadpipe data, use MCP.

Open source

The MCP server is MIT licensed and open source:

npm: npx -y @leadpipe/mcp
GitHub: github.com/leadpipe-com/mcp
SDK: github.com/leadpipe-com/sdk / npm install @leadpipe/client

To get an API key, sign up for Leadpipe (free trial, 500 leads, no credit card).