Meta Ads integration
An agent can run the campaign. Nothing spends until you say so.
Ask most people whether they would let an AI spend their ad budget and the answer is no. That is the right answer to the question they heard, which is an agent spending unsupervised while you find out later. This is not that. Campaigns are drafted, projected and held until you approve them, and the numbers tell you afterwards whether it worked.
Meta's own MCP server, authorized by you
Meta Ads connects through Meta's official MCP server rather than an integration we built on top of the ad API. You authorize it with your own Meta business account and the ad accounts you choose, and your workflows can then use it as an action step. That means the permissions are Meta's, revocable by you at any time, and the connection behaves the same way as the other tools your workflows reach into.
Campaign work that runs from a workflow
Describe it in plain language. The agent does the assembly and the reporting, and the approval gate stays where you put it.
"Build a campaign for the new product, show me the projected spend, and wait for my approval"
The agent assembles the campaign, the audience and the budget and stops. You see what it intends to spend before anything runs.
"Every Monday, pull last week's performance into a summary for the team"
You read the results instead of assembling them. The numbers come from the ad account rather than from a screenshot somebody pasted.
"If a campaign's cost per result goes above our threshold, tell me and pause it"
The check that only happens when somebody remembers to look now happens on every run.
"Build an audience from the customers who messaged us on WhatsApp this quarter"
The people already talking to you are the ones worth reaching. That connection is only possible because the messages and the ad account are in the same place.
Ads are verifiable, which is why an agent can do the work
The output can be checked
The place agents earned trust first was coding, because a test suite tells you whether the work is right. Ads have the same property: the numbers say whether a campaign is working, so you are checking against something objective rather than taking it on faith.
The approval gate is the point
Nothing spends until you approve it. The risk people are objecting to when they say no to this is unsupervised spend, and that is the part we removed rather than the part we ask you to accept.
Reversible by design
A campaign that is not working can be paused or pulled. Unlike an email that has been sent, an ad spend decision stays adjustable while it runs.
The same place as the customer
The person who messaged you and the campaign that reached them are handled in one app, so the audience you build can come from the conversations you actually had.
An assistant for campaign work, not an autopilot
This is not a tool that decides your strategy or reallocates budget on its own while you sleep. It builds what you describe, reports what happened, and waits at the gates you set. If you want an agent that spends without asking, this is the wrong product, and we would rather say that here than have you find out with a live budget.
Run the campaign from the same place the customers are
Connect Meta Ads and a workflow can build, adjust and report on campaigns, holding at your approval before anything spends.
Free during beta · No credit card required