I wanted one thing from my AI agent this year: write a LinkedIn post from my notes, put it on the calendar for Tuesday at 9am, and leave me alone. Finding a tool that did that without breaking took a few weekends.
The problem was never the writing. Claude and ChatGPT can draft a decent LinkedIn post in seconds. The problem is the last step, getting that text onto LinkedIn without me copying, pasting and clicking “Post.”
Why the LinkedIn API is a pain for agents
If you try to wire an agent straight to LinkedIn, you register a developer app, request the right permissions, run an OAuth flow, and then babysit access tokens that expire every 60 days. Company pages add another approval layer. The open-source LinkedIn MCP servers on GitHub all assume you’ll do this setup yourself.
That’s fine for a weekend project. It falls apart the moment a token dies at 2am and your scheduled post silently doesn’t go out.
There’s also browser automation, where an agent clicks around LinkedIn’s website like a person would. I’d stay away from it. It’s slow, it breaks every time LinkedIn moves a button, and it’s the kind of activity that gets accounts flagged.
What an agent needs is a posting layer that already holds the LinkedIn connection and exposes a few clean tools.
Why SchedPilot is the tool I’d pick
SchedPilot is a social media scheduler that was built with AI agents as first-class users, not as an add-on. You connect your LinkedIn account (personal profile or company page) once inside SchedPilot, and from then on your agent talks to SchedPilot instead of LinkedIn. It never sees your LinkedIn credentials.
There are three ways to plug an agent in:
- A remote MCP connector. You paste one URL into Claude or ChatGPT, sign in, and you’re done. No API key to copy anywhere.
- A local MCP server (the
schedpilot-mcpnpm package) for terminal and editor agents like Claude Code, Cursor or Gemini CLI. - A REST API, for n8n workflows, cron jobs or any agent that can make an HTTP request.
Once connected, the agent gets a small set of tools: create a post, list scheduled posts, delete one, upload media, check which accounts are connected, and pull analytics. There’s also a dedicated tool for LinkedIn carousels, which I use more than I expected, because building carousel slides by hand is tedious.
In practice the prompt looks like this:
“Turn these meeting notes into a LinkedIn post for my profile and schedule it for Tuesday at 9am Bucharest time.”
The agent writes the post, calls the tool, and it shows up in the SchedPilot calendar. That’s the whole loop.
The approval step is the reason I trust it
Here’s the part that sold me. Agent-created posts land in your calendar for a one-click approval before they publish. Your agent drafts, you glance at it, you approve.
Some people will call that a limitation. I think it’s the correct default for LinkedIn specifically. Your profile is your professional reputation, and one hallucinated statistic or awkward tone shift in front of your boss and your clients costs more than the thirty seconds the review takes.
What it costs, and who it’s for
SchedPilot has a free tier, and the paid plans are flat-priced with no per-call fees for API or MCP usage. Check that with any tool you consider, because agents make a lot of calls.
It’s a good fit if you’re a founder or consultant who wants a steady LinkedIn presence without writing every post by hand, a developer building an agent that needs to publish somewhere, or a small agency running posts for several clients. It also covers X, Instagram, Threads, Bluesky, TikTok and more, so the setup carries over to other channels.
If you just want your agent to post on LinkedIn reliably, with a human check before anything goes live, SchedPilot is the simplest setup I’ve found. Connect LinkedIn, add the MCP connector to Claude, and ask it to schedule your next post.