Automate

A LinkedIn connector for Claude that posts from the chat

LinkedIn is the one platform where the agent can also read and answer the comments, and the only one that takes a PDF as a post.

One value to paste
https://api.antwork.io/mcp

Claude connects to Antwork as a custom connector: one URL in the Add custom connector dialog, then a sign-in on Antwork's own consent page. The same connector works in Claude Desktop once you are signed in.

Claude Pro, Max, Team, or Enterprise required. Custom connectors are not available on the free Claude plan.

The short answer

Yes. Claude can post to LinkedIn through Antwork, a Model Context Protocol (MCP) connector. You paste one URL into Claude's Add custom connector dialog, approve LinkedIn once on LinkedIn's own OAuth screen, and from then on Claude drafts, schedules and publishes posts to your personal LinkedIn profile or to a Company Page you manage, from inside the conversation. Antwork calls LinkedIn's official API with the scopes you granted (w_member_social for your profile, w_organization_social for Company Pages), so there is no API key to generate, no browser extension to install, and no logged-in browser session being driven on your behalf.

Antwork is a publishing connector, not a prospecting one. It drafts, schedules and publishes posts, reads the engagement on posts you own, and replies to comments underneath them. It does not search or scrape profiles, enrich leads, send connection requests, or send direct messages.

LinkedIn MCP servers compared: which post through the official API, and which drive your browserChecked against the product on 3 October 2026

Connect it

0:59 · Captions EN / ES

Every screen is a real recording of claude.ai and Antwork, made on 26 September 2026. The chat list in the sidebar is blurred for privacy; only the cursor, zooms and captions are added.

  1. Click Connect and we'll open Claude's Add custom connector dialog with the MCP URL copied to your clipboard. In claude.ai you can also get there yourself: Customize, then Connectors, then Add, then Add custom connector.

Is there a native Claude connector for LinkedIn?

Anthropic does not publish a first-party LinkedIn connector, but Claude supports custom connectors over MCP and Antwork is one. You add it from Claude's own connector settings and its tools become available like any other: no browser extension, no API key, and no intermediate workflow to maintain. These are the approaches you are actually choosing between, and they differ in what each one leaves you holding.

  • An analytics connectorPulls LinkedIn numbers into the chat so you can ask about followers, impressions or ad spend in plain language. It reads. Publishing is not part of what it does.
  • A prospecting or outreach serverProfile lookup, enrichment and drafted messages, aimed at finding people rather than posting. These generally ask you to paste a vendor API key into the connector, so the vendor sits between Claude and LinkedIn holding a key and whatever it gathered about the people you searched for.
  • A general integration toolkitOne MCP endpoint that fronts hundreds of apps. LinkedIn is one entry in a catalogue, and the tools you get are the generic ones that catalogue exposes.
  • A workflow builderYou build and maintain a flow, then trigger it. Claude is a node inside the automation rather than the thing driving it, and every change means editing the flow.
  • A browser extensionDrives LinkedIn inside your own logged-in tab. Nothing to authorize, and nothing authorized either: it is your session being automated, which is the arrangement LinkedIn's User Agreement is most explicit about.
  • Antwork, a publishing connectorDrafting, scheduling and publishing to LinkedIn are first-class tools, running against LinkedIn's official API on scopes you granted, with the platform's real limits enforced server-side. Claude drives them from the conversation. No vendor API key.

What you can say once LinkedIn is connected

“Draft a LinkedIn post about what we shipped this week, in my voice, and show me before it goes out.”

“Turn this PDF into a LinkedIn carousel post and schedule it for Tuesday morning.”

“What did my last five LinkedIn posts do? Show me impressions and engagement.”

“Someone commented on my last post. Read the thread and draft a reply.”

You see it before it goes out

Antwork does not answer with a wall of text. It renders real, clickable panels inside Claude — the same interface you would get in a dashboard, in the conversation you are already in. No other social tool does this.

  • 01A post card with the draft, editable in place, with the publish and schedule buttons on it
  • 02A month calendar of everything scheduled across every connected account
  • 03Your media library as a thumbnail grid you can attach from
  • 04A connections panel with a Connect button per platform
  • 05A table of recent posts you can click into
Rendered in the chat
Draft · LinkedIn268 / 3,000

Shipped something small this week that fixed a big thing. Our token refresh ran daily but checked LinkedIn against a five minute expiry window, so it could only ever catch a token that happened to die inside those five minutes. It never did. Accounts just went quiet.

PublishScheduleedit in place

What is different about LinkedIn

Only here

It can reply to comments

LinkedIn is the only platform Antwork implements commenting for. The agent can read the thread under a published post and reply in the same voice it wrote the post in. On every other platform it publishes and stops.

A PDF is a carousel

LinkedIn accepts a document post — a PDF up to 100 MB that readers swipe through in the feed. Attach one and the agent posts it as a carousel. No other platform Antwork supports takes a document at all.

A document can ride with images, video cannot

You can attach a document alongside images, but images and video can never share a post. Ask for both and the agent has to split them, so it will tell you before it drafts rather than after it fails.

Long actually works here

Three thousand characters is roughly 500 words, and LinkedIn is the one place where using them is not a mistake. The agent drafts to the length the format rewards instead of trimming everything to a hook.

Questions

Can Claude connect to LinkedIn?
Yes. Antwork is the connector: you add one URL to Claude, approve LinkedIn once over OAuth, and from then on Claude drafts, schedules and publishes to your account from the chat itself. No API key, no browser extension, no copying between tabs.
Is there a native Claude connector for LinkedIn?
Anthropic does not publish a first-party LinkedIn connector, but Claude supports custom connectors over MCP and Antwork is one. You add it from Claude's own connector settings and its tools become available like any other: no browser extension, no API key, and no intermediate workflow to maintain.
Is there a LinkedIn connector for Claude?
Not an official one. Neither Anthropic nor LinkedIn ships a LinkedIn connector for Claude. Claude supports custom connectors over the Model Context Protocol (MCP), and Antwork is a third-party LinkedIn connector built on it: you add Antwork's server URL in Claude's connector settings, approve LinkedIn once on LinkedIn's own consent screen, and Claude can then draft, schedule and publish posts to your profile or Company Page through LinkedIn's official API.
How do I give Claude access to my LinkedIn profile?
Through a connector, never a password. Add Antwork as a custom connector in Claude (Settings, then Connectors, then Add custom connector, and paste the server URL), then ask Claude to connect LinkedIn. LinkedIn's own consent screen opens; once you approve it, Claude can publish to your profile or to a Page you manage. What it gets is permission to post and to read the engagement on your own posts, not your inbox or anyone else's profile.
Will connecting Claude get my LinkedIn account restricted?
Antwork is not the kind of access LinkedIn restricts, and LinkedIn does not block Claude specifically. What it restricts is automated access to the site itself: the LinkedIn User Agreement prohibits scraping and driving linkedin.com with bots, extensions or browser automation, and that is what most Claude-to-LinkedIn setups are built on. Antwork never touches the website. It calls LinkedIn's official API with a token you granted on LinkedIn's own consent screen, so what LinkedIn sees is an authorized application using the scopes you approved: w_member_social to post as you, w_organization_social to post as a Page you manage.
Why is the connector not found in Claude?
Almost always the plan. Custom connectors are a paid Claude feature, so on the free plan the Add custom connector option is absent rather than broken. On Pro, Max, Team or Enterprise it sits under Settings, then Connectors, then Add custom connector. If the dialog is there but Antwork's tools never appear after you add it, the OAuth approval did not finish: open the connector again and complete the browser consent screen.
Can Claude read my LinkedIn messages or other people's profiles?
No. Antwork requests only the scopes it needs to publish on your behalf and to read engagement on your own posts. It never receives your LinkedIn password, it has no messaging scope so it cannot open your inbox or your direct messages, and it cannot look up anyone else's profile. You can revoke the connection from LinkedIn at any time and the token stops working immediately.
How is this different from Zapier, n8n or an integration toolkit?
What drives the action. In a workflow builder, Claude is a node inside an automation you maintain. With Antwork, drafting, scheduling and publishing to LinkedIn are tools Claude operates directly in the conversation, and the platform's limits are enforced server-side rather than being whatever the flow happens to carry.
Do I need an API key to connect Claude?
No. Antwork connects over OAuth — you approve a consent screen in the browser and you are done. There is no key to generate or paste into a URL, which is how most social MCP servers handle it.
Do I need a paid account?
No. Your first two connected accounts are free forever, with unlimited posts and workspaces, and there is no plan gate on MCP access.
Will Claude post without showing me first?
Not unless you ask it to. Drafting and publishing are two separate steps: the agent writes, renders the post card, and only publishes when you confirm.
Does the writing sound like AI?
Antwork hands the agent that account's recent posts, along with how each performed, and the agent writes to match them. On LinkedIn that includes writing to the length the format rewards, up to 3,000 characters.

Never leave Claude to post again.

One command, one consent screen, and LinkedIn is something you can just ask for.

Connect Claude to LinkedIn