Meta Banned OpenClaw, but I Plugged It into My Ad Account

There’s a piece of news making a lot of noise in tech circles lately.

Reportedly, Meta executives handed down a strict order to their teams: nobody may install OpenClaw on company devices, and anyone who does faces dismissal.

The first reaction from quite a few friends in advertising: “Peter, Meta has banned OpenClaw, is it dead?”

No, it’s not dead. In fact, I finished connecting it last night.

1. What exactly did Meta ban

Meta’s ban targets internal employees installing OpenClaw on company devices. It started when an employee used it to handle a work mailbox; mid-task the AI forgot the “no action without approval” instruction and wiped the inbox clean, and in the end the person had to rush over and pull the network cable.

Out of internal network security concerns, Meta promptly ordered the ban, with dismissal for violators.

This has nothing to do with us advertisers. We go through Meta’s official Marketing API, which has clear permission boundaries; here OpenClaw is only a reader of data and a sender of instructions. They’re two completely different scenarios.

2. Meta wants to do this too

While the outside world debates whether OpenClaw is dangerous, Meta had earlier announced a $2 billion acquisition of an AI Agent company, with plans to embed it into Ads Manager. The official line: help advertisers automatically generate reports, analyze data and optimize campaign recommendations.

The platform itself stepping in to do this is certainly not on the same scale as a third-party tool with API authorization.

3. What I did

In my last article I shared the results of an experiment running PA (PropellerAds) with OpenClaw: $339 spent, $457 revenue, 34.9% ROI. Many friends asked whether FB could be done the same way.

Yesterday I finished that connection. Specifically, I used an open-source CLI tool and completed the app registration and permission token setup on the Meta developer platform.

With the whole process done, OpenClaw can now read my ad account data and has the basic ability to execute actions.

Right now it’s at the communication stage, not yet the stage of handing over full control. Before I formally let it run on its own, I need to design the rule logic rigorously enough. After all, I don’t want a rerun of that deleted-inbox story, just with my ad budget in the leading role.

4. Some thoughts

The same piece of news, and different people see completely different things. In Meta banning OpenClaw, most people see danger, instability, don’t touch it.

Look at it from another angle: an AI Agent that can operate a computer on its own is already powerful enough for a tech giant to feel threatened. So can it operate my ad account for me?

Because this lands in a field I know well, its value is easy to see. For a traditional media buyer, the same information is a blind spot, and the opportunity may slide right past.

The way I see it, there’s no harm in trying. Just design the boundaries before you start, then play.