Schedule LinkedIn Posts From Claude or ChatGPT (Free)

Most scheduling tools ask you to open a dashboard, paste your text into a box, click a calendar, pick a time from a dropdown, and press save. That is five steps for something you could say in one sentence. If you are already writing the post inside Claude or ChatGPT, moving it into a separate tab just to set a time is the slowest part of the whole job.

Conversational scheduling removes that step. You finish writing, you say when, and it is booked. This page explains exactly how that works, what you still have to approve, and the one limitation you genuinely need to plan around.

How do I schedule a LinkedIn post from Claude or ChatGPT?

You add a connector once, then you just ask. A connector here means an MCP server — MCP is the Model Context Protocol, an open standard that lets an AI assistant use outside tools. Kalovio is one of those servers, aimed at LinkedIn, and it is free during its public beta.

After the one-time setup, the flow looks like this:

• You write the post in the chat, the usual way — drafting, rewriting, cutting the waffle out of the opening line.
• You say something like “schedule that for Monday at 9am my time”, or “post it tomorrow morning”, or “put it out on the 14th at 7:30”.
• The assistant calls the schedule_post tool, and you get a preview back: the exact text, the exact time, and anything attached.
• You reply “yes”. Only then is it booked.

That is the whole loop. There is no dashboard, no calendar widget, no second login. The setup step is a normal OAuth sign-in — you approve the connection on LinkedIn’s own page and can revoke it from your LinkedIn settings whenever you like. Nothing asks for your password or a copied browser cookie.

Does it get my time zone right?

Yes, and this is worth being specific about, because “9am” is meaningless without knowing 9am where.

You set your time zone once, and after that every time you say is read in your local clock. Ask for “Tuesday at 8am” from Mumbai and it publishes at 8am in Mumbai, not 8am UTC. Behind the scenes the time is converted and stored in a single absolute reference so it cannot drift, but you never have to think in offsets.

You can also set a standing weekly pattern — your usual posting slots, say Tuesday and Thursday at 9am — using set_posting_schedule. Once that exists, you can hand over several drafts at once with queue_posts and they get dropped into the next free slots in order, up to 30 posts in a single request. That is the closest thing here to a content calendar, and it still lives entirely in conversation.

What do I see before anything gets scheduled?

A preview, always. This is not a nicety — it is a safety gate built into the server, and it cannot be skipped.

Scheduling is a two-step action. The first call returns the preview plus a signed token — a short cryptographic stamp covering the exact content: the text, the publish time, any image, any mentions, any tracked links. Nothing is booked until a second call arrives carrying your confirmation and a token that still matches. Change one character and the old token no longer fits.

The practical reason for that design is prompt injection: hidden instructions buried in a web page or a document the model reads, telling it to post something. A gate that requires you to approve the exact wording is the defence. It also quietly protects you from the ordinary version of the same problem — the assistant misunderstanding you and booking the wrong thing.

So the preview is doing two jobs at once. Read it as a proofread as well as a permission slip.

Will it still post if my laptop is closed?

Yes. This is the difference between a scheduler and a reminder.

Once a post is booked, it lives on the server, not in your chat window and not on your machine. Kalovio runs on Cloudflare Workers with one timer per user — a Durable Object alarm, which is simply a server-side clock that wakes up at the right moment and does the job. At that moment the server decrypts your LinkedIn token in memory, sends the post through LinkedIn’s API, and records the result.

You can close the chat, close the laptop, be on a plane, or be asleep. None of it matters. You do not need Claude or ChatGPT open for a scheduled post to publish, and you do not need to be signed in anywhere.

How do I see what is scheduled, and cancel it?

By asking, in the same conversation or a completely new one.

• “What have I got scheduled?” calls list_scheduled_posts and returns everything waiting, with its time and its id.
• “Cancel the one on Thursday” calls cancel_scheduled_post and removes it before it goes out.
• “What are my posting slots?” calls get_posting_schedule and shows the weekly pattern.

The queue is tied to your account, not to a chat session. Book something from Claude on your desktop today, and you can list and cancel it from ChatGPT on a different machine next week. There is no state trapped in a conversation you have since deleted.

Editing is a cancel-and-rebook, not an in-place edit. If you want to change the wording of a scheduled post, cancel it and schedule the new version — which also means the new wording goes through the preview gate, as it should.

Can I schedule images and carousels, or only text?

Text and images, yes. Documents, not yet.

One row needs explaining properly.

Attachments have to be uploaded before scheduling, because neither Claude nor ChatGPT can hand a chat attachment to a connector — that is a limitation of both apps, not of the server. You upload through Kalovio’s upload panel or a browser link, which takes about as long as dragging a file into a message. Files can be up to 20 MB each, 100 per account, kept for 30 days.

What is the catch with scheduling far ahead?

There is one, it is real, and it is the thing most worth remembering on this page: your LinkedIn connection expires roughly every 60 days.

LinkedIn issues member access tokens with about a 60-day life, and it does not give self-serve apps a refresh token — the thing that would let a tool quietly renew in the background. Partner-tier apps get that. Self-serve apps do not. So every couple of months you reconnect, which is one click through LinkedIn’s own consent screen.

The consequence for scheduling is straightforward. You can book a post up to a year ahead, but if your connection lapses before that date arrives, the post cannot publish when the timer fires. The server is not able to renew the permission on its own, and it should not be able to — that is the same property that makes the whole thing revocable by you.

Three habits make this a non-issue:

• Ask “is my LinkedIn still connected?” now and then. The linkedin_account_status tool answers in a second and tells you where you stand.
• Reconnect before a long gap. Going away for a month, or booking a run of posts into next quarter? Run reconnect_linkedin first and start the 60 days fresh.
• Keep the far-future queue short. Scheduling two or three weeks ahead is comfortable. Scheduling nine months of content in one sitting is asking a 60-day permission to cover a 270-day plan.

Any tool built on LinkedIn’s self-serve products has this constraint. The difference is whether it tells you before your post silently fails to appear.

Is this a replacement for Buffer or Hootsuite?

No, and it is not trying to be. Those are multi-network dashboards with team approvals, analytics, and a proper calendar view. If that is what you need, use them.

This is a narrower thing: LinkedIn only, no dashboard, driven by conversation. It suits one specific person — someone who already writes their posts inside an AI assistant and wants the scheduling step to disappear rather than move to another tab. The queue is real, the timing is server-side, and the approval is explicit. It is not a content marketing suite.

Two honest limits worth knowing before you commit. First, you cannot get LinkedIn engagement analytics through it — likes, comments and impressions need LinkedIn’s partner-gated Community Management API, which requires a registered legal company and app review. Link clicks are tracked, through short links, but that is a different measurement. Second, it cannot read posts you did not publish through it, because the permission for reading a member’s wider post history is closed to self-serve apps.

What does it cost?

Nothing at the moment. Kalovio is in free public beta with no card required. Pricing may change later, and it would be dishonest to pretend otherwise, but there is no trial clock running and no upsell in the flow today.

One cost note that is not ours: Claude works on the free plan. ChatGPT needs Developer Mode, which is only on the paid plans and only on the web, and the connector has to be switched on per conversation from the tools menu. That is an OpenAI restriction, not a Kalovio one.