I’m Building My Own Social Media Automation System With AI

I’m building a social media automation system called PostCue. This is not another SaaS product I plan to sell. It is a tool I want to use myself.

The idea is pretty simple. I want one place where I can create content, schedule it, see what is coming next, and understand what happened after I posted it. I do not want to keep paying for expensive software when I may be able to build the exact workflow I need.

What is PostCue?

PostCue is my attempt to run more of my social media process on my own. The plan is to connect it to Instagram and TikTok so I can manage the work from one private dashboard.

I am building it for the way I work. That means I do not need every feature that belongs in a giant social media platform. I need the few things that will actually help me create consistently, see what is working, and spend less time moving between tools.

What I want it to do

If I get the full version working, PostCue will let me:

• connect my Instagram and TikTok accounts;
• create and schedule posts;
• see everything on a content calendar;
• automatically repost selected content;
• see views, likes, saves, shares, and comments;
• track total engagement without pulling the numbers together by hand.

That is the vision. I am not pretending all of it works yet.

Right now, I have a mockup and a blocker

The interface is taking shape. I can see what the dashboard, calendar, post creator, and analytics could look like. But right now it is still a mockup of the system, not a fully connected social media tool.

The blocker is my Meta developer account. I need that account working before PostCue can connect to Instagram through Meta’s API. An API is the controlled connection that lets one piece of software talk to another. Without that connection, I can design the experience, but I cannot honestly say the Instagram publishing and analytics work.

This is part of building too. Ai can help me create the screens and write the code, but it cannot skip another company’s account approval, permissions, or platform rules.

Why build this instead of buying it?

The goal is control. I want to run the system myself, shape it around my own process, and avoid another expensive monthly software bill if I can.

Here is where I want to be completely transparent. My first thought was that ChatGPT could build the app, Cloudflare could run it for free while my usage was low, and I could avoid most of the normal software cost. That may be possible, but it is not quite as simple as I first made it sound.

ChatGPT Sites can turn a prompt or an existing project into a hosted website or lightweight app. Sites is the service doing the hosting in that setup, and Sites has its own plan-based usage limits. It is still in beta. If a Site uses too much, OpenAI says that can affect whether it stays public or whether I can add more storage. That means I should not describe the hosting as guaranteed to be free forever.

Deploying an app directly to my own Cloudflare account would be a different setup. Cloudflare does have generous free allowances for Workers, its D1 database, and R2 file storage. For a small personal tool with light usage, those allowances could be enough. But “could be enough” is not the same as “everything will run for free.” A social media tool can store a lot of video, run scheduled jobs, make repeated API calls, and process analytics. Those are exactly the things that can push usage up.

There are other possible costs too. I already pay for the Ai tools helping me build it. I may eventually need more hosting, storage, processing, or another service to make the Instagram and TikTok connections reliable. Right now, I am testing how far I can get with the tools and allowances I already have.

That is part of what I want to share. If the cost stays close to zero, I will show you why. If it does not, I will show you what created the cost and whether I think it was worth it.

You can check the current details in the official ChatGPT Sites documentation ↗, Cloudflare Workers pricing ↗, Cloudflare D1 pricing ↗, and Cloudflare R2 pricing ↗.

The bigger lesson is not that everyone should build PostCue. It is that you can look at the software you pay for, identify the small part you actually use, and ask whether Ai can help you build a personal version around your own workflow.

The desktop version is the easy part

One thing I keep noticing with these builds is that Ai does a pretty good job with the desktop version first. Then I open the same thing on my phone and find the strange spacing, awkward line breaks, buttons that feel too big, and sections that simply do not look right.

The easiest way I have found to fix it is not technical at all:

1. Open the page on your phone.Use the product the way a real person will use it.
2. Take a screenshot.Capture the exact section that looks wrong.
3. Send it straight to the chat.The image gives Ai the same visual context you have.
4. Talk through what you see.I usually use voice and explain each issue as specifically as I can.
5. Fix one section at a time.I keep going until the mobile version feels intentional, not like a smaller desktop screen.

I will say things like, “This heading breaks in the wrong place,” “There is too much empty space under this card,” or “This button is touching the edge of the screen.” The more specific I am, the better the fix usually is.

What happens next

I am still working at it. The next meaningful step is getting my Meta developer account working and connecting the mockup to real Instagram data. After that, I can find out whether the scheduling, analytics, and reposting parts work the way I imagined them.

When that happens, I will update this article and share what worked, what did not, and what I had to change.

A few quick questions