How to let ChatGPT post on LinkedIn (and what works today)

The short answer

Yes, ChatGPT can post to LinkedIn through Finchreach. Finchreach’s MCP server supports OAuth sign-in, so you add it as a custom connector in ChatGPT’s developer mode, sign in to Finchreach and approve. No API key to paste.

You have three options:

• Add Finchreach as a custom connector in the ChatGPT app. Developer mode needs a Plus, Pro, Business, Enterprise or Edu plan.
• Use OpenAI’s Codex CLI to draft and submit posts through Finchreach, with an API key or OAuth. Same OpenAI models, different app.
• Keep ChatGPT as your writer and paste the final text into LinkedIn yourself. It’s low-tech, but there’s no account risk beyond normal use.

Whichever route you pick, Finchreach keeps one rule: every post, comment, message or invitation your assistant creates waits in your approval queue. Nothing goes live on LinkedIn until you approve it.

How do I connect ChatGPT to LinkedIn?

MCP (Model Context Protocol) is a standard plug that lets AI apps use outside tools. ChatGPT supports MCP through apps and connectors: in developer mode you can add a custom connector by pasting a server URL and choosing OAuth for authentication. Finchreach’s server at https://mcp.finchreach.com/mcp supports OAuth, so it connects directly.

1. Create a Finchreach account at finchreach.com/sign-up and connect your LinkedIn account under Accounts.
2. In ChatGPT, open Settings → Apps & Connectors → Advanced settings and turn on Developer mode.
3. Click Create. Name it Finchreach, set the MCP server URL to https://mcp.finchreach.com/mcp and Authentication to OAuth, then create it.
4. Sign in to Finchreach when asked. On the consent page, choose the LinkedIn account and what ChatGPT may do, and approve.
5. In a chat, open the + menu, go to Developer mode and turn Finchreach on. Then ask it to list your LinkedIn accounts to check the connection.

Developer mode needs a Plus, Pro, Business, Enterprise or Edu plan. On Business, Enterprise and Edu, an admin may need to allow it first.

Details on each plan and screen are in our ChatGPT guide, along with what we checked and when.

Can I post to LinkedIn with Codex instead?

Yes. If you prefer a terminal, the Codex CLI runs OpenAI models, speaks MCP, and can send a bearer token taken from an environment variable. Setup takes about five minutes:

1. Create a Finchreach account at finchreach.com/sign-up and connect your LinkedIn account under Accounts.
2. Create an API key under API keys. Copy it now; it’s shown only once.
3. Set the key in your shell: export FINCHREACH_API_KEY=lk_...
4. Add the server to ~/.codex/config.toml (snippet below), or run codex mcp add finchreach --url https://mcp.finchreach.com/mcp and add the token line afterwards.
5. Start Codex and ask it to list your LinkedIn accounts. If you see your account, you’re connected.
6. Ask for a post. Codex submits it, Finchreach queues it, and you approve or reject it under Approvals.

[mcp_servers.finchreach]
url = "https://mcp.finchreach.com/mcp"
bearer_token_env_var = "FINCHREACH_API_KEY"

bearer_token_env_var tells Codex to send Authorization: Bearer <value>, so your key never sits in the config file. Our Codex guide covers project-level config and troubleshooting, and the docs list every tool.

What can I ask my assistant to do on LinkedIn?

Once ChatGPT or another assistant is connected, you talk to it as you normally would. Here are prompts that map to real Finchreach tools:

Ask your assistant“Look at my last few LinkedIn notifications and my profile headline. Then draft a short post about what I learned shipping our onboarding redesign this week. Keep it under 150 words, no hashtags, and submit it for my approval.”

Ask your assistant“Read the top posts in my LinkedIn feed today and tell me which three are worth a thoughtful comment from me. Draft a comment for each, but don’t submit anything until I pick one.”

Ask your assistant“I just finished this blog post (pasted below). Write a LinkedIn version that leads with the main takeaway, then submit it as a text post for approval.”

Behind those prompts, the assistant uses tools such as linkedin_get_profile, linkedin_get_notifications, linkedin_get_feed and linkedin_create_post. Posts are text only today: no images, video, scheduling or company pages. If you ask for an image post, a well-behaved assistant should tell you it can’t attach one.

Can I keep using ChatGPT for writing and something else for posting?

Yes, and for many people that’s the most comfortable setup. ChatGPT is where your custom instructions, past chats and saved style notes already live. Use it for what it does well: finding the angle, tightening the hook, trying three versions of the first line.

When the text is final, hand it to a connected assistant with a plain instruction:

Ask your assistant“Submit this exact text as a LinkedIn post, without changing a word: [paste]”

Or skip the second assistant and paste it into LinkedIn yourself. Automation pays off most when the assistant needs context it can only get from LinkedIn: reading your feed, checking who commented, replying in a conversation, or seeing who sent you an invitation. Publishing a post you already wrote is the step that gains least from it.

What happens after my assistant submits a post?

Finchreach doesn’t publish straight away. The tool call returns a pending_approval status and the post shows up in your approval queue with the exact text that would go live. From there:

• Approve it, and Finchreach publishes it on LinkedIn as you.
• Reject it, and nothing is sent. Your assistant can check the status with linkedin_get_action and offer a new draft.
• Leave it, and it expires instead of going out at some surprise moment later.

The same applies to every write: comments, reactions, direct messages and connection invitations. Reading (feed, notifications, messages, search) happens right away, because it doesn’t put anything out in your name.

The approval step does more than catch typos. Assistants sometimes misread a request, and anything they read, such as a post in your feed, can contain text written to steer them. A person between the draft and LinkedIn means nothing goes out under your name that you haven’t seen.

Is it safe to let an AI post on my LinkedIn?

It’s safer than handing over your password, but it isn’t risk-free, and we’d rather say so up front.

Automating a personal LinkedIn account carries risk under LinkedIn’s User Agreement, which prohibits unauthorized automated methods for actions such as messaging, commenting and posting. Finchreach’s approval queue and your own pacing lower that risk; they don’t remove it. No tool can honestly promise your account won’t be restricted.

What helps in practice: approve each post yourself, post at a normal human pace, keep the voice your own, and don’t let any tool run bulk outreach. We go deeper in how to automate LinkedIn with AI safely and in our honest look at LinkedIn automation without getting banned.

On access itself: when you approve a connector, you choose the LinkedIn account and scopes, and API keys have scopes too. A key with only linkedin:read can browse but can’t submit anything, and the default key can’t approve its own actions, so approval stays with you in the dashboard.

ChatGPT and LinkedIn: frequently asked questions

Does ChatGPT need an API key? No. The ChatGPT connector uses OAuth: you sign in to Finchreach and approve. API keys are for clients that send a header, such as Codex CLI or n8n. You can revoke either kind of access in the Finchreach dashboard.

Does Finchreach have its own AI writer? No. Finchreach is the bridge between your assistant and LinkedIn. The writing comes from whichever model you use, which keeps your prompts, style and model choice in your hands.

Can it schedule posts for later? Not today. Posts publish when you approve them. If timing matters, approve at the moment you want the post to go out.

Can it post images or to a company page? No. Text posts on your personal profile only.

Which assistants work without extra steps? ChatGPT and claude.ai through OAuth sign-in, and anything that can send a custom header to a remote MCP server: Claude Code, Codex CLI, Cursor, VS Code with GitHub Copilot, Windsurf, n8n and others. See the Claude Code integration, the Claude guide and our explainer on what a LinkedIn MCP server is.

Where do I start? Create an account, connect LinkedIn, make a key, and follow the quickstart.