Claude + LinkedIn MCP: Find and Message High-Intent Leads in Real Time

LinkNav is an MCP connector that gives Claude direct access to LinkedIn. Once connected, you can tell Claude to find leads, run campaigns, and pull campaign analytics - all from a prompt in the Claude chat window. No manual searching, no spreadsheet exports, no switching tools.

How to connect it

Setup in 3 steps

What Claude can do once connected

The full capability set

Instead of pulling a static Apollo list and blasting it, this approach finds people who have posted or commented on LinkedIn in the last 24 hours and are already talking about the problem you solve. These leads are warm before you say a word.

The prompt to find active leads

Copy and adapt this

Why active in the last 24 hours matters

Timing is the variable most teams ignore

Once Claude finds your leads, you can do one of two things: generate a personalised message to send manually, or enrol them into an automated campaign. Here's how both work.

Option 1: Generate a personalised message and send it yourself

The prompt

Option 2: Enrol them in a campaign

Building the campaign

If/then conversation logic

After a campaign runs, ask Claude to analyse the performance. It pulls the data from LinkNav and gives you a breakdown of what worked, what didn't, and what to carry into the next campaign - including a geographic and ICP breakdown of who actually connected.

The prompt

How to run it

What the analysis actually tells you

The most important metric is why

What metrics to look at

Instead of finding leads manually and messaging them one by one, you can build a Claude skill that does the entire workflow automatically: scrape leads, filter out the wrong ones, score the rest by intent, create a campaign, and queue it for your review. You run it once. Claude handles the rest.

What the skill does end to end

The full workflow

What you need before building it

Two connectors

The prompt to build the skill

What to give Claude

What Claude builds

Running the skill and reviewing the output

What happens when you run it

Token cost note