Can ChatGPT post to LinkedIn?
By Ryan Melling, founder of WordPush
· 5 min read
ChatGPT can publish to LinkedIn through a custom connector, but only on paid plans and only with developer mode on. What's required, and how to set it up.
ChatGPT can post to LinkedIn, but not on its own and not on the free tier. You add a custom MCP connector that holds your LinkedIn authorisation, turn on developer mode to do it, and from then on you can ask ChatGPT to draft, illustrate and publish a post without leaving the chat.
It is a slightly longer path than the equivalent in Claude, entirely because of where OpenAI has put the setting. Here is what is actually required.
What you need before you start
Three things, and the plan requirement catches most people out.
- A paid ChatGPT plan. Plus, Pro, Business, Enterprise or Edu. Custom MCP connectors are not available on the free tier at all, and no amount of prompting works around it.
- Developer mode, switched on. Custom connectors live behind it.
- A hosted connector. ChatGPT connects over remote HTTPS only. There is no stdio option, so a server running on your laptop is not reachable — it has to be something someone hosts, or something you have deployed publicly.
On Business and Enterprise plans there is a fourth: your administrator may have to permit developer mode organisation-wide first. If the switch simply is not there, that is almost always why, and it is a conversation with IT rather than a bug.
Turning on developer mode
OpenAI has moved this more than once, so the menu names below are current as of September 2026 rather than permanent.
- Open Settings → Apps & Connectors.
- Switch on Developer mode.
ChatGPT will warn you that this is elevated risk. That warning is accurate and worth reading rather than clicking past: developer mode lets you add MCP servers OpenAI has not reviewed, which means the trust decision moves from OpenAI to you. For a connector from a company you are already paying, that is a reasonable trade. For a server you found in a GitHub search, think harder.
Adding the connector
Still in Apps & Connectors:
- Choose Create.
- Give it a name and a short description — these are just labels for you.
- Paste the server URL. This must be the full MCP endpoint path. For WordPush:
https://app.wordpush.co.uk/mcp/wordpush - Set authentication to OAuth.
- Confirm the warning and create it.
- Approve the sign-in when prompted.
The single most common setup failure is the URL. People paste the product’s website address rather than the MCP endpoint, get an unhelpful error, and conclude the connector is broken. The endpoint almost always has a path on the end. If a vendor has not told you what theirs is, that is a bad sign about the rest of the integration.
The full walkthrough with the current screens is on the ChatGPT setup page.
Set the permissions, then test on something harmless
Once connected, decide what the connector is allowed to do before you give it anything real.
Leave the read-only tools freely available — reading your brand context or checking a render status cannot hurt you. Require approval for anything that writes, and keep publish on approval permanently. Publishing is the only action that is instantly public and cannot be undone from inside the tool; you would have to go and delete the post from LinkedIn, by which time it has been seen.
Then test with something you would not mind existing. Ask it to save a draft rather than publish one. Check the draft appears where you expect. That confirms the whole chain — ChatGPT to connector, connector to your account — before anything is at stake.
ChatGPT or Claude?
If you have not committed to either, Claude is the easier setup for this specific job: custom connectors work on every plan including Free, there is no developer mode switch, and the per-tool permission controls are clearer.
ChatGPT’s advantages are elsewhere — if it is where you already work and where your context lives, the extra two minutes of setup is irrelevant. The finished experience is much the same.
What is not worth doing is switching assistants over this. Once connected they call the same tools on the same server and produce the same posts.
There is a fuller version of this comparison, including Claude Code and Cursor, in can Claude post to LinkedIn.
The part that is actually hard
Connecting ChatGPT to LinkedIn solves publishing friction. It does not solve the thing that stops most people posting, which is not knowing what to say.
And there is a second-order problem worth naming: if ChatGPT writes your post, it sounds like ChatGPT. Anyone who spends time on LinkedIn now recognises the cadence — the triads, the “it’s not X, it’s Y”, the tidy little summary at the end. A post that reads as machine-written gets less engagement than one that does not, so the speed you gained on drafting is partly given back at the other end.
Two ways to avoid it. Give ChatGPT enough of your own writing that it can match you, and redo that work in every new conversation. Or let the connector do the drafting with a voice model built from you, so ChatGPT decides what to say and the tool decides how you say it. WordPush’s write_post tool works the second way — the tool reference sets out which tools do what.
Either way: add the specific. A real number, a named client, a thing that actually happened this week. That is what no model can supply and what makes a post sound like a person.
In short
Yes, ChatGPT can post to LinkedIn. You need a paid plan, developer mode, and a hosted connector with an OAuth login rather than one that wants your session cookie. Set publish to ask-first, test on a draft, and be deliberate about whose voice the post ends up in.
WordPush runs an MCP server on the official LinkedIn API, included in the plan. Setup guides for ChatGPT, Claude, Claude Code and Cursor.