How to schedule LinkedIn posts from Claude

You can schedule LinkedIn posts by asking an AI assistant, in the same conversation where you wrote them. Connect a tool that exposes a scheduling capability, say “schedule it for Tuesday at 9am”, and it goes into a queue that publishes whether or not you ever open the assistant again.

The interesting part is not that it is possible. It is that it changes how you plan a week’s posting, and it introduces one failure mode worth knowing about before it bites you.

How scheduling actually works

Your assistant does not hold the post and wait. It has no persistent process, no clock, and no ability to do anything between conversations. If scheduling depended on the assistant remembering, it would never fire.

What happens instead: the assistant calls a scheduling tool on a connector, passing the post and a time. That connector is a service that runs continuously. It writes the post to its own calendar and publishes at the appointed moment through LinkedIn’s official API. The assistant’s involvement ends the second the tool returns “scheduled”.

This matters practically. You can close the chat, close the laptop, and the post still goes out. It also means the post is now visible in that tool’s calendar, so you can move or cancel it there without going back through the assistant.

Time zones: the mistake everyone makes once

Your assistant does not know what time zone you are in, and it will not ask.

When you say “schedule it for 9am”, that time is interpreted by the scheduling tool, using whatever zone your account is set to. If your account was created with a default of UTC and you are in Los Angeles, “9am” is 1am your time — and the post goes out to an empty feed.

The fix is thirty seconds and permanent: check the time zone in your publishing tool’s settings once. After that, “Tuesday at 9am” means what you think it means, every time.

Two habits that help while you are getting used to it:

• Ask it to confirm. “Schedule for Tuesday 9am and tell me the exact time it will publish.” A good connector returns the scheduled time in its response, so you can see immediately if it is wrong.
• Think about your audience’s zone, not yours. If you sell to the US from the UK, 9am your time is the middle of the night for most of the people you want to reach. This is a strategy question the tool cannot answer for you.

The thing that makes this genuinely better than a calendar UI

Scheduling one post from an assistant is marginally faster than scheduling it in a web app. Scheduling five is dramatically faster, and that is the case worth making.

In a normal scheduling interface, each post is an isolated act: open composer, paste, pick a date, save, repeat. The context resets every time. In a conversation it does not. You can work like this:

“Here are four things that happened this week. Draft a post for each, in my voice.”

“Make the second one shorter and cut the last line.”

“Good. Schedule them Tuesday, Wednesday, Thursday and Friday, 9am each, and space the two about hiring so they aren’t next to each other.”

That last instruction is the one a calendar UI cannot take. You are asking for a judgement about the sequence, holding all four posts in view at once. Twenty minutes on a Monday and the week is done.

This is also the honest argument for batching in general, which we made at more length in how to post on LinkedIn consistently: consistency fails when it depends on you having something to say at the moment you have time to say it. Separating the deciding from the publishing is what fixes that, and an assistant is a good place to do the deciding.

Setting it up

The mechanics are the same in any MCP client:

1. Add the connector. In Claude: Customize → Connectors → + → Add custom connector, then paste the server address. In ChatGPT it is Settings → Apps & Connectors, with developer mode on first.
2. Approve the sign-in.
3. Set permissions. Scheduling is a write action, so decide whether you want to approve each one. Most people allow scheduling freely and require approval for publish now — scheduled posts can be cancelled before they fire, immediate ones cannot.
4. Ask.

Per-client walkthroughs: Claude, ChatGPT, Claude Code, Cursor.

Scheduling and publishing are different permissions, and should be

It is worth separating these deliberately.

Scheduling is recoverable. The post sits in a queue. If the assistant misunderstood you, or the draft was worse than you thought, you can open the calendar and delete it before anything happens.

Publishing is not. Once it is out, it is in feeds. Deleting it afterwards does not undo the people who saw it, and on LinkedIn a deleted-and-reposted item tends to perform worse anyway.

So the sensible configuration is: let it schedule, make it ask before it publishes. You keep the speed on the common action and the safety on the irreversible one. WordPush exposes these as two separate tools — schedule_post and publish_post_now — precisely so they can carry different permissions. The full list is on the tool reference.

What about timing?

The question underneath “when should I schedule this” is usually “when is the best time to post on LinkedIn”, and the honest answer is less exciting than the listicles suggest.

Weekday mornings work reasonably well. Tuesday to Thursday tends to edge out Monday and Friday. Beyond that, the variance between your audience and the averages in any study is larger than the effect you are optimising for.

What reliably matters more: posting at all, regularly, with something worth reading. A good post on a Wednesday afternoon beats a mediocre one at the statistically optimal hour, every time. If you have limited attention to spend, spend it on the post rather than the slot. How the LinkedIn algorithm works covers what actually drives reach.

Getting started

WordPush runs an MCP server that schedules and publishes through LinkedIn’s official API, to your profile or a company page, with the calendar visible in the app. It is included in the single plan, along with the voice model that writes the drafts.

Start with the Claude setup guide, or see how the connector works.