Meta Ads MCP: querying Facebook and Instagram from Claude or ChatGPT

For Meta the connector returns the same families of data as for other platforms: connected accounts, performance over a period, campaigns and budgets. Creatives are added — the ads themselves — which matters on a platform where ad fatigue shows at creative level, not at campaign level.

• “Which Meta campaigns saw their cost per acquisition double in two weeks?”
• “Compare my Meta and Google Ads spend over the quarter.”
• “Which creatives are still running while no longer converting?”

Today the permissions Meta grants at connection time cover reading. A Meta recommendation therefore says “apply this yourself” rather than offering a button that would fail on click.

In practice, the assistant can tell you which campaign to cut and why; for now the cut happens in Meta Ads Manager. Write permission has been requested from Meta: creating and optimising your Meta campaigns from SaveAds is being activated.

Two differences, and they matter. First, search terms: they do not exist on Meta, where targeting does not run through queries — so wasted spend is a question of audience and creative, not of keyword.

Second, execution: on Google Ads, what the assistant recommends can be applied from SaveAds with safeguards and rollback. On Meta, the loop stops at the recommendation for now; applying from SaveAds is being activated.

The point of the connector is not Meta alone: it is being able to ask a question that crosses platforms. Every connected account comes back through the same tool, with its currency, so the assistant can put side by side spend that normally lives in four separate interfaces.