Skip to main content
The AI Agent is a conversational assistant connected to your Meta Ads account. You ask questions and give instructions in plain language. It reads your campaign data, explains what is happening, and can make changes to your account, always with your approval first. Where: sidebar → Agent. Cost: free to chat (some actions like image generation cost 1 credit). Plans: all (model tiers vary by plan). Recommended reading order: start with Permissions, then Approvals, then Model tiers, and finally Pro features.

Get started

1

Connect a Meta account

Open the Agent and select an ad account. If you have not connected one, follow Connect Meta Ads.
2

Ask a question

Type something like “How are my campaigns doing this week?” or “Which ad set has the worst CPC?” The agent already knows your brand profile, so it never asks basic questions about your business.
3

Approve actions

When the agent proposes a change, it shows an approval card. Nothing happens until you click Approve. See The approval system.

What it can do

Read and analyze (automatic, no approval needed): campaign overviews, performance (spend, impressions, clicks, CTR, CPC, CPM, reach, frequency), ad set and ad data, account-level insights, and your brand context. Modify your account (requires approval): pause or resume campaigns and ads, and change campaign budgets. Generate creatives (requires approval, costs credits): generate new ad images (1 credit each) with the same engine as the Ad Generator, create variations of an existing ad (1 credit each) like the Ad Copier, and upload creatives to Meta (free). Uploaded ads are always created paused.

Examples

  • “How are my campaigns doing this week?”
  • “Which ad set has the worst CPC?”
  • “Pause the campaign that’s spending the most with the lowest return.”
  • “Generate 3 new ad images for my landscaping business.”

Learn more

Model tiers

Lite, Raddivus, and Pro, and when to use each.

Permissions

The five toggles that control what the agent can see and do.

Approvals

How the agent asks before it changes anything.

Pro features

Extended thinking, task planning, and sub-agent delegation.
Last modified on June 6, 2026