The New OpenAI Agent Builder Wasn’t Made For You

In partnership with

DEEP DIVE
The New OpenAI Agent Builder Wasn’t Made For You

The internet called OpenAI’s new Agent Builder an “n8n killer” the day it dropped.

And I get it (kinda).

This really could have been a Zapier-meets-GPT moment, and many wanted this so bad that they tried to imagine it into existence…

Because… it’s not.

Why? Because here’s the truth: this thing wasn’t built for you.

Not for small business owners.

Not for indie builders.

And definitely not for the people who actually use automation tools to run their businesses.

OpenAI is basically Apple 2.0

They hand you the tools, cheer you on while you build, and then — once you find something people actually want — they launch their own version.

It’s like the App Store all over again, except this time, they own the models instead of the devices.

That’s what AgentKit feels like.

So what are people doing with it right now?

Let’s break it down:

• Some folks are forcing automations into it that are 3x easier to build in n8n or Zapier.

• Others think they cracked the code by connecting it back to Zapier to access 4,000+ apps. Which is like fixing a Bluetooth printer by plugging in a USB cable.

• And then you’ve got the casual crowd calling it a “less complicated n8n,” which translates to: “It doesn’t really work, but hey — it’s honest work.”

Here’s the thing:

I’ve played with it. I’ve watched people push its limits.

And so far, I haven’t seen anything that justifies the hype, at least not for your average AI enthusiast or small business.

Because again: it wasn’t made for you.

It wasn’t made to compete with Zapier, n8n, or Make.

It was made to keep all of us inside OpenAI’s ecosystem.

What’s really going on

This is a distribution play.

OpenAI watched AI automations blow up across other tools and said,

“Cool idea. We’ll take it from here.”

AgentKit is step one in making sure your workflows, your agents, your little AI side hustles, live inside ChatGPT, not outside it.

Right now, it’s clunky and dev-heavy.

But give it a few updates and my guess is that you’ll be able to call these “agents” (spoiler: still not real agents) directly from the ChatGPT chat.

If you zoom out, AgentKit isn’t a competitor to no-code tools: it’s more like CustomGPTs 2.0.

And one more thing:

This tool wasn’t made for non-technical folks.

It shipped alongside the SDK for a reason. It’s built for devs who want to prototype logic before shipping it inside their own apps.

Just look at the UI:

If you showed this to a non-techy person, they’d think it’s the Matrix.

So yeah — call me underwhelmed.

AgentKit is cool in theory, but it’s not the automation revolution Twitter promised you.

But hey that’s just my take on it, I’m curious to know what you think of this new feature: