On 16 July 2026, Meta opened its ads MCP server to any developer with a Meta app. In plain terms, that means an AI assistant can now create campaigns, edit ad sets, build audiences, pull performance data and run A/B tests inside a Meta ad account, through conversation rather than through the ads manager interface.
That’s a significant release and it’s been covered mostly as either a threat or a miracle. It’s neither. We’ve been building this kind of infrastructure for our own accounts for a while, so this post is the practical version: what it does, what it can’t do, and what it means depending on whether you run your own ads or pay somebody else to.
#Key summary
- Meta opened its ads MCP server to all developers on 16 July 2026
- AI agents can now create and manage campaigns, audiences, catalogues and experiments through natural language
- Agencies managing accounts for others need Advanced Access through Meta’s app review process
- The tool set is rolling out gradually rather than arriving complete
- This raises the floor on execution speed, it doesn’t raise the ceiling on strategy
- The competitive advantage moves further towards offer, creative and judgement
#What an MCP server is, without the jargon
Model Context Protocol is a standard for letting AI assistants talk to other software. Before it existed, connecting an AI tool to a platform like Meta meant somebody writing custom integration code for that specific pairing, then maintaining it every time either side changed.
MCP replaces that with a common connector. Meta builds one server describing what its ads system can do, and any AI application that speaks the protocol can use it. The integration work drops to roughly nothing.
That’s the whole idea. It’s plumbing. But plumbing determines what gets built on top of it, which is why this matters more than a feature release.
#What it can actually do
The capabilities Meta has exposed are broader than most coverage suggests. An AI agent connected to the ads MCP server can create, edit and delete campaigns, ad sets and ads. It can manage custom audiences. It can handle catalogue work, including creating catalogues, adding products and managing product sets.
On the reporting side it can query performance metrics across flexible date ranges, covering spend, impressions, click-through rate and return on ad spend. It can access signal diagnostics, which matters if you’ve ever tried to work out why your pixel data looks wrong.
It can also create and manage A/B tests and conversion lift studies, and search Meta’s Business Help Centre.
Two practical notes. If you’re an agency or partner managing data on behalf of other businesses, you need to submit for review and request Advanced Access. And Meta has said the full tool set is rolling out gradually, so what’s available today isn’t the finished picture.
#What it absolutely cannot do
This is the section that matters, and it’s the one you won’t find in the excited coverage.
It has no judgement about your business. It doesn’t know your margins. It doesn’t know that the leads from one campaign close at 40% and the leads from another close at 4%. It doesn’t know that half your enquiries are tyre-kickers who found you through a discount angle you regret running. Optimising towards a conversion number without knowing what those conversions are worth is how accounts spend efficiently on the wrong thing.
It doesn’t know what you can deliver. We’ve watched businesses scale campaigns beautifully into a capacity wall they couldn’t service, then damage their reputation working through the backlog. No amount of automated optimisation catches that, because the constraint lives in the workshop, not the ad account.
It can’t write a better offer. The largest performance gains in almost every account we touch come from changing what’s being offered or who it’s aimed at, not from restructuring campaigns. That’s a business decision that requires knowing things the ad platform has no access to.
Speed multiplies direction, including the wrong one. An agent that can restructure your account in ninety seconds can also restructure it badly in ninety seconds. Faster execution is only an advantage once the direction is right.
The honest summary is that this technology is very good at the work that was always mechanical, and no help at all with the work that was always hard.
#If you run your own Meta ads
The floor is rising. Competent execution, which used to be a real differentiator for the business owner who’d put in the hours to learn the platform, is becoming widely available.
That’s mostly good news for you. The tedious parts of self-managing ads (the naming conventions, the bulk edits, the pulling of numbers into a spreadsheet on a Sunday night) are exactly what this handles well.
But it means your advantage has to come from somewhere else. Your advantage is that you know your customers, your margins and your capacity better than any agency or agent ever will. Lean into that. Spend the time this frees up on your offer and your creative, not on learning more platform mechanics.
One warning. Conversational access makes it very easy to make sweeping changes without a record of what changed or why. If you go down this path, keep a change log. Accounts get destroyed by twenty small undocumented changes far more often than by one big one.
#If you pay someone to run your Meta ads
Fair question to ask them: what are you doing with this?
There’s a version of agency work that was always mostly execution, and that version is in genuine trouble. If your agency’s value has been building campaigns and sending you a report, the tooling is closing in on that.
The version that survives is the one doing the things the agent can’t: understanding your economics, choosing the strategy, making the creative, and exercising judgement about when the numbers are telling you something misleading.
You’re not looking for an agency that avoids this technology, and you’re not looking for one that treats it as the whole answer. You’re looking for one that uses it to spend less time on mechanics and more on your business.
Reasonable questions: What have you automated? What do you deliberately keep manual, and why? How would I know if the automation made a bad call?
#What we’ve built
We run our own agent infrastructure across the accounts we manage, including automated review cycles that flag creative fatigue and structural issues before they cost money, and reporting that assembles itself rather than being rebuilt monthly by hand.
We built it because the alternative was spending our best hours on work that didn’t need a human. What it hasn’t done is replace the part of the job that clients actually pay for, which is deciding what to do.
We’re saying that plainly because there’s a lot of noise in this space right now, and the businesses selling “fully autonomous advertising” are, in our experience, describing something that doesn’t exist yet. See our Meta Ads training if you want to build the judgement that agents can’t replace.
#Frequently asked questions
#Do I need to do anything about this?
No. This is developer infrastructure, not a change to your ad account. Nothing happens to your campaigns because Meta released it.
#Can I connect my own AI assistant to my Meta ads?
Technically yes, if you hold a Meta app and go through the developer dashboard. Practically, it’s a developer task rather than a business owner one. Managing accounts on behalf of other businesses also requires Advanced Access through app review.
#Is this going to replace ad agencies?
It will replace the parts of agency work that were purely mechanical. Strategy, creative and commercial judgement aren’t in scope, because the agent has no access to what makes those decisions correct for your business.
#Is it safe to let an AI make changes to a live ad account?
With guardrails, yes. Without them, no more or less than letting an inexperienced person do it. Spend limits, change logs and a human reviewing anything structural are the sensible minimum.
#Does Google have an equivalent?
The direction of travel across every major platform is the same. Assume this becomes normal rather than exceptional over the next year or two.
#Where this goes
The prediction we’re comfortable making is unexciting. Execution becomes cheap and abundant. Everyone gets access to competent campaign mechanics. The gap between a well-run account and a badly-run one narrows.
And then the advantage sits entirely where it always really sat: in having a better offer, better creative, and a clearer understanding of what your customers are actually worth.
It’s a threat to anyone whose value was doing the easy part slowly.
This guide is written by Hedgehog, a DIY digital marketing consultancy specialising in small and medium businesses in Australia. We offer digital marketing consulting, coaching and training.