LinkedIn MCP Server: What Exists and What Is Safe to Use

Search for "LinkedIn MCP server" and the top result drives a real browser with your LinkedIn session in it. Its own FAQ says LinkedIn can restrict or ban accounts that use automated tools, and it cannot publish a post. Most people searching that phrase want the opposite: an AI agent that posts to LinkedIn without putting the account at risk.

The short version: LinkedIn does not run an MCP server of its own. The servers you will find fall into three groups: ones that drive your logged-in browser, ones that call LinkedIn's official API, and hosted tools that sit in between. Only the official-API kind can post without breaking LinkedIn's User Agreement, and it needs one permission, w_member_social. Check that before you connect anything to your profile.

Does LinkedIn have an official MCP server?

No. LinkedIn publishes REST APIs, not an MCP server, so every "LinkedIn MCP" you find is built by someone else on top of something else. That matters because MCP is only the doorway your agent walks through. What happens behind it, a real API call or a script clicking buttons in a browser, decides whether your account is safe.

MCP (Model Context Protocol) is the standard way agents such as Claude, ChatGPT and Cursor call outside tools. A server exposes a few named tools; the agent calls them. An MCP server for LinkedIn is just a server whose tools end up touching LinkedIn. The question is how.

The three kinds of LinkedIn MCP servers

Kind	How it reaches LinkedIn	Can it post?	Account risk
Browser-session servers	Reuses your logged-in browser or cookies and acts in the page	Usually not; built to read profiles, jobs and messages	High: automated access is what LinkedIn's terms forbid
Scraping platforms	Third-party scrapers behind an MCP front end	No; they extract data	High for the data you pull; not a posting tool
Official-API servers	LinkedIn's own sign-in page, then LinkedIn's REST API	Yes, with w_member_social	Low: this is the route LinkedIn built for apps

The first two groups are popular because they can read things the official API keeps private, like other people's profiles and your inbox. That is also why they are risky. Section 8.2 of LinkedIn's User Agreement lists, among things you agree not to do, using "bots or other unauthorized automated methods" to access the service, including to create, comment on or share posts, and using scripts, crawlers or browser plugins to scrape it.

What LinkedIn's API lets an MCP server do

LinkedIn's Posts API is the official way to publish. A server posting for you calls POST https://api.linkedin.com/rest/posts with a token that carries w_member_social, the permission to "post, comment, and like posts on behalf of an authenticated member." Posting for a company Page needs w_organization_social and a Page admin role, and LinkedIn reviews apps before they get Page access.

Here is what that looks like in our own connector, so you have something concrete to compare against:

• Sign-in: LinkedIn's own consent page, asking for openid, profile and w_member_social. Nothing else.
• Publishing: POST /rest/posts with the LinkedIn-Version header LinkedIn requires. Skryvo sends one image per post through the Images API.
• Checking: after publishing, we read the post back with GET /rest/posts/{id}. A post only shows as published once LinkedIn returns it.
• Pages: our Page app is in LinkedIn's Community Management review, so Page posting is built but not switched on yet.

Notice what is missing: reading your feed, your messages or other people's profiles. Those permissions are restricted, and a posting tool does not need them. If a "LinkedIn MCP" offers them, it is probably not using the official API.

Is there a Claude MCP for LinkedIn?

Yes, through a connector that posts to LinkedIn for you. Skryvo is listed in Claude's connector directory: choose Connect, sign in to Skryvo, approve access, and connect your LinkedIn profile in Skryvo once. After that you can ask Claude things like:

• "Write a LinkedIn post about our launch week and schedule it for Tuesday at 9am."
• "What is scheduled on LinkedIn this week?"
• "Cancel the LinkedIn post scheduled for Friday."

Claude drafts the post in your brand's voice, shows it to you, and only schedules it after you say yes. The same connection also covers X, Bluesky, Mastodon, Pinterest, Telegram, Discord, Slack and your blog, so one request can become a LinkedIn post and an X thread. The steps for each agent are on our Claude page.

Can ChatGPT or Cursor post to LinkedIn?

Not on their own. Both can call a remote MCP server, and that is how they post. In ChatGPT, add the server under Plugins as a custom MCP server; in Cursor, add it under Tools & MCP in the desktop app. The server address is the same for every agent, and each one signs you in through the same consent page. ChatGPT steps and Cursor steps are written out separately.

How to check a LinkedIn MCP server before you connect it

1. Where do you sign in? It should be LinkedIn's own consent page. A server that asks for your LinkedIn password, a cookie or a browser extension is driving your session.
2. Which permissions does it ask for? Posting needs w_member_social. Read the list on LinkedIn's consent screen before you approve.
3. Does it publish through /rest/posts? Open-source servers show this in their code; hosted ones should say so in their docs.
4. Can you revoke it? You should be able to remove access in LinkedIn's settings and in the tool itself, and access should stop at once.
5. Do you need your own LinkedIn developer app? Several open-source servers ask you to register one. That works, but it is your app to maintain and keep approved.

Where an MCP server stops

An official-API server will not read your feed, send connection requests or reply to messages, because the posting permission does not cover those. It also will not post to a company Page until LinkedIn approves the app for Pages. If you need those things, you are back to tools that automate your session, with the risk that comes with it.

For most people the posting part is what saves the time anyway. Connect a server that signs in through LinkedIn, ask your agent for this week's posts, and read each one before you say yes.