Building a LinkedIn MCP Server: 16 Tools for Automated Social Media

Today I deployed a custom LinkedIn MCP server to our AI Hub infrastructure. It gives Claude direct access to LinkedIn's API for posting, engagement, and analytics—all through natural conversation.

This post you're reading? The LinkedIn announcement was written and published by Claude using this exact system.

Why Build a LinkedIn MCP?

I run two brands that need consistent social presence: Millyweb Development (technical/developer content) and RevenueFirst.AI (service business consulting). Posting to three LinkedIn accounts (personal + two company pages) manually was eating 30-45 minutes daily.

The solution: Give Claude the ability to post directly. One conversation can now create tailored content for all three audiences and publish it immediately.

The 16 Tools

The MCP server exposes 16 tools covering the complete LinkedIn workflow:

Publishing

post_to_personal

Post to personal profile

post_to_millyweb

Post to Millyweb company page

post_to_revenuefirst

Post to RevenueFirst company page

delete_post

Remove a post by URN

Engagement

add_comment

Comment on any post

get_comments

Retrieve post comments

delete_comment

Remove a comment

add_reaction

Like, celebrate, support, etc.

remove_reaction

Undo a reaction

Analytics

get_org_followers

Follower count and growth

get_org_page_stats

Page views and visitors

get_post_stats

Impressions, clicks, engagement

get_org_posts

Profile

get_profile

Authenticated user info

get_connection_count

1st-degree connections

get_linkedin_urns

Configured account URNs

Technical Architecture

The LinkedIn MCP is a Node.js HTTP server (~200 lines) that wraps LinkedIn's REST API in MCP protocol format. It runs as a Docker container on our VPS2 workload server.

Request Flow

Claude.ai

→

OAuth Gateway

→

Tool Filter

→

Context Forge

→

LinkedIn MCP

The server is registered with Context Forge (IBM's MCP gateway) alongside MCPJungle and other direct integrations. This brings our total tool count to 339 tools from 21 servers.

        Key Numbers
        16 LinkedIn-specific tools
339 total tools in ecosystem
95% token reduction via semantic filtering
~200 lines of code
$0 additional cost (self-hosted on existing VPS)

      

LinkedIn API Setup

Getting LinkedIn API access requires creating an app in the LinkedIn Developer Portal:

Create app with "Sign In with LinkedIn" product
Request "w_member_social" permission for posting
For company pages: verify as admin and add "w_organization_social"
Generate access token (expires in 60 days, plan for refresh)

The URNs for your profile and company pages can be found via LinkedIn's API or through the page admin settings.

What This Enables

With LinkedIn MCP integrated, I can now:

"Post this to all three LinkedIn accounts with tailored messaging"
"Show me engagement stats for yesterday's posts"
"What's the follower growth trend for Millyweb?"
"Reply to comments on my latest post"

All through natural conversation with Claude. No switching tabs, no copy-pasting, no scheduling tools.

Building in Public

This is part of our ongoing AI Hub infrastructure build—creating enterprise-grade automation on a $40/month VPS budget. Everything documented, everything open.

The code pattern is reusable for any REST API you want to expose to AI assistants. Twitter/X, Facebook, Instagram—same approach, different endpoints.

Want to build your own MCP servers?

Check out our MCP infrastructure guide or reach out to discuss custom implementations.

Learn More About MCP