Meta Ads AI Connectors: how to connect your Meta account to an AI agent safely
Since 29 April 2026 Meta lets you connect your ad account to an AI agent such as Claude or ChatGPT through its own MCP server. The agent doesn't just read reports — it creates and edits campaigns. Here is what it can do, how to connect it and what to restrict so you don't lose budget.
What did Meta launch?
On 29 April 2026 Meta opened a beta of Meta Ads AI Connectors — an MCP server and a command-line tool (Ads CLI). They create a secure, Meta-authenticated connection between your ad account and an AI agent. The MCP route needs no developer credentials or code. On 16 July Meta added connecting any AI app through your own developer app, plus ads MCP server rules: whoever has full control of the business portfolio can stop the agent from, for example, changing budgets or the catalog.
What can the agent do?
| What the agent can do | What it means for you |
|---|---|
| Reporting: campaign performance, insights | Questions in plain language instead of manual exports |
| Management: create and edit campaigns, ad sets and ads | The agent changes a live account — mistakes cost money |
| Catalog: create, add products, fix feed issues | Find and fix products that don't show in ads |
| Signal diagnostics: health and quality of conversion data | A quick check of the pixel and Conversions API |
The key difference from ordinary reporting integrations: the agent can edit campaigns, not just read them. So connect it gradually and with limits.
How to connect without risking your budget
- Start with reporting: for the first weeks ask the agent only for analysis and recommendations, and make changes yourself.
- Set ads MCP server rules: if you have full control of the business portfolio, block budget and catalog changes until you trust the agent.
- Review every change before confirming: significant edits send an ad set back into the learning phase, so frequent agent edits can hurt results.
- Reconcile numbers with the account: always state the date range and attribution window in your prompt, or the agent and Ads Manager will show different results.
What else should you know?
- It's an open beta: some tools may reach your account gradually.
- Ads CLI needs developer access and a command line; MCP only needs a Meta login.
- Agencies working with client accounts through their own app must pass Meta's app review.
- Meta hasn't announced pricing.
Metriverra shows Meta spend and results daily next to your other channels, so it's easy to check what the agent changed and how it affected sales.
In short
- Since 29.04.2026 a Meta ad account can be connected to an AI agent via an MCP server, in open beta.
- The agent handles reporting, campaign management, catalogs and signal diagnostics.
- Since 16.07 there are MCP server rules — restrict budgets and catalog.
- Start with read-only use and review changes by hand.
- Always give the agent the date range and attribution window.
Sources: Meta for Business — Introducing Meta Ads AI Connectors · Meta Business Help Center — Meta ads AI connectors · Meta for Developers — Introducing Ads CLI · PPC Land — Meta opens ads MCP to any app
