LinkedIn MCP: Give Your AI Agent Access to Your Professional Network

If you're searching for a LinkedIn MCP, you want the same thing most people do: let your AI assistant see your professional network. Who you know. Who they know. Who can get you in the door somewhere.

Here's the honest state of that, and the fastest way to get it working.

What is a LinkedIn MCP?

MCP (Model Context Protocol) is the standard that lets AI assistants like Claude and ChatGPT connect to outside tools and data. A LinkedIn MCP server gives your assistant access to your professional network, so it can answer questions like "who can introduce me to someone at Stripe?" instead of guessing.

Is there an official LinkedIn MCP server?

No. LinkedIn does not publish one, and its official API gives third-party apps almost no access to your connection data. Every LinkedIn MCP you find today is unofficial. But there's still a way to access the data that matters, your network graph, and that's with Draftboard.

Here are your options.

Option 1: an open-source LinkedIn MCP server

There are open-source LinkedIn MCP servers on GitHub. Most work the same way: you run the server on your machine, hand it your LinkedIn login or session cookies, and the agent drives your account directly.

For personal experiments (pull a profile, read a page) they're fine. Two problems show up fast:

1. The agent is the automation. It clicks around your logged-in session, and it doesn't know LinkedIn's rate limits. Your account carries all of that risk.
2. You get raw profile data, not answers. Knowing you have 34 mutual connections with someone doesn't tell you which of them actually knows the person well enough to make an intro.

Option 2: Draftboard's hosted MCP

Draftboard is a warm-intro platform. It maps the paths from your network (and your team's) to the people you want to reach, and scores each relationship 0 to 100 with the reason behind the score: worked together at Google for two years, ending three years ago. That data is what the MCP serves.

The difference in architecture matters: your assistant never touches LinkedIn. Draftboard's Chrome extension does the mapping separately, at a deliberately conservative pace. The assistant just queries the results.

The connector has full read and write access:

• Ask for paths. "Who's my warmest path into Stripe?" comes back with connectors, scores, and why each one is strong.
• Check connections. "Does anyone on my team know Jane Doe?"
• Add targets. Paste 40 prospects in chat and mapping starts in the background.
• Track intros. Mark requests sent, completed, or declined without leaving the conversation.

How to connect it (about two minutes)

1. Create a Draftboard account (you get 50 free credits) and stay signed in.
2. In your assistant, add a custom connector with one address: `https://mcp.draftboard.com`
3. Approve the connection when your browser opens. No API key to create or paste.

This works in Claude (Desktop and Claude Code), ChatGPT (developer mode, paid plans), Codex CLI, and Grok (custom connectors on grok.com). Exact steps for each client are at draftboard.com/mcp.

One expectation to set: results are not instant. Draftboard maps your network in real time, so a big target list takes a day or two to finish. Ask your assistant to check back.

What does it cost?

Usage-based, no subscription, no seats. About 70% of the paths Draftboard finds are free. Only quality paths (e.g., work or school overlap) cost credits, capped at 3 credits per prospect, so one prospect never costs more than $0.90. New accounts start with 50 free credits.

One honest limit: scoring runs on public work and school history. A close friendship that never shows on LinkedIn can score low. The trade is worth it, since it means very few false positives, but know it's there.

Try it

Sign up, paste `mcp.draftboard.com` into your assistant, and ask: "who's my warmest path into [company you're trying to crack]?"