LinkedIn MCP server for posting from Claude or ChatGPT

You tell Claude “take the report PDF and publish it as a document on the company page Tuesday at 9 a.m., titled ‘Five numbers from our industry’”. The document is scheduled on your LinkedIn with its title and shows up on your team’s calendar. That’s HeyMark’s LinkedIn MCP server: it connects your profile or your company page to Claude, ChatGPT or any MCP client that supports OAuth, so you can post and reply to comments without leaving the conversation.

One clarification first: searching “LinkedIn MCP” turns up servers built to scrape profiles and job listings. This page isn’t about that. HeyMark works with your own LinkedIn presence, not with other people’s data, and MCP access is free on every plan.

What your assistant can do on LinkedIn

Posts and documents

It creates, schedules and publishes posts and documents, which are LinkedIn’s PDF carousels, with the title the document displays. It can mention people and companies in the text, and post to your profile or to a company page. Since HeyMark pulls in what you already published, your assistant can read your earlier posts and write new ones in the same voice.

Plan with your team

The idea, the copy and the file live in the same post, with a status, a date and an owner. Your assistant can assign the piece to someone, leave a private note for the team, move it to another status and create a link to share the calendar with dates, time zone and permissions. If a client signs off on the copy, HeyMark has an approval link on every plan. Say an agency runs a dental clinic’s LinkedIn: it leaves the month’s document as a draft, the clinic approves it through the link, and the assistant schedules it. How the rest of your networks get scheduled is in schedule social media posts.

Comments

It reads the comments on your posts, drafts replies and posts them once you approve. It also creates keyword rules: when someone comments the word, it replies publicly. On LinkedIn the rules don’t send DMs; that’s for Instagram, Facebook and TikTok.

Metrics

LinkedIn is part of HeyMark’s analytics overview: followers, views and interactions against the previous period, plus metrics for each post. If someone else needs the numbers, your assistant can turn on a public link to the report.

What needs your approval

Publishing right away, replying to a comment, creating or changing a rule, creating a calendar link and turning on the public report link show you a preview to approve first. Creating drafts, assigning people and scheduling run right away, so check the date, time and copy before you ask it to schedule. What gets published is your call.

Prompts for LinkedIn

Create a document for my brand's LinkedIn with the PDF I pick in HeyMark's file picker.
Document title: [title]. Post text: [text].
Show me the draft and wait for my confirmation before scheduling it for Tuesday at 9 a.m.

Read my last 10 LinkedIn posts and notice how I write: length, structure and sign-off.
In that voice, write a post about [topic] and leave it as a draft, not scheduled.

Go through the comments on my 3 latest LinkedIn posts.
Tell me which ones ask a question and draft a reply for each.
Don't post any reply until I approve it.

On the webinar post, create an automation: if someone comments REGISTER, reply publicly with "Thanks, here's the link: [link]".
Show me the rule before turning it on.

How did my LinkedIn do in the last 30 days compared with the 30 before?
I want followers, views and interactions, plus the 3 posts with the most interactions.

How to connect it

1. In HeyMark, open My accounts in your brand and connect LinkedIn.
2. Add the URL https://mcp.heymark.ai to your assistant: in Claude as a custom connector (guide for Claude) and in ChatGPT with developer mode (guide for ChatGPT). The Claude Code and Codex commands are in the MCP documentation.
3. Approve the access when HeyMark opens in your browser, then ask your assistant which LinkedIn account it sees.

Limits

• It doesn’t scrape profiles, contacts or job listings.
• It isn’t an official LinkedIn MCP server. It uses the account you connected in HeyMark, with your permissions.
• LinkedIn automations reply publicly; they don’t send DMs.
• Each brand connects one LinkedIn account.
• Scheduling has no preview: a scheduled post goes out on time.
• LinkedIn accepting the request doesn’t mean the post is live. Ask your assistant to check the actual status.
• It doesn’t manage paid ads, and there’s no public REST API yet.