# Interactive Cards

The eleven interactive cards the AI Copilot renders in the conversation, what you can do inside each, and what happens in assistants that do not support them.

Canonical: https://adsap.ai/docs/guides/ai-copilot/interactive-cards
Updated: 2026-09-13

In assistants that support **MCP Apps**, which is Claude today, some Adsap tools do more than reply with text. They render a card in the conversation that you can read, sort, tick and act inside.

Nothing to set up. Cards appear on their own wherever the assistant supports them.

## Why cards rather than text

A table of thirty campaigns is unreadable as a paragraph, and pausing three of them by describing which three is worse. Cards exist for the cases where looking and clicking beats reading and dictating.

The assistant knows what you did inside a card. When you tick rows or change a budget, that goes back into the conversation, so you can say "launch into that one" without naming it again.

## The eleven cards

### Ad preview

Ask for a preview of an existing ad and you get Meta's own render of it, not an approximation. Tabs for Facebook Feed and Instagram Reels, a button for every other placement, and for video ads an animated clip of it playing.

Open this card before you launch anything. Believing your ad looks right and seeing that it does are two different things.

### Ad Library

A competitor's ad from Meta's Ad Library: the creative at full size plus its copy as text. Appears when you look up one ad from an Ad Library search.

### Upload progress

Appears while creatives transfer from Google Drive or Dropbox and while videos encode. A progress bar that moves, so you do not have to ask whether the upload finished.

### Performance snapshot

The KPI row, a daily trend chart and a ranked table of campaigns or ads, from one request. Covers the whole answer to "how did last week go", which is why one request per account and date range is enough.

### Account health

A traffic-light scorecard: account status, payment health, business verification, pixel health, and the review status of your ads. The card to open when something has stopped working and you do not yet know what.

### Google keywords

Sortable tables of keyword ideas and of the searches that triggered your ads, with negative-keyword candidates highlighted. Go here to find what wastes money and what deserves promoting.

### Manage campaigns

The most useful one. A table of your campaigns or ad sets where you can:

- **Tick rows** to select them
- **Pause and activate** without leaving the card
- **Edit budgets** in place

A ticked row is a choice the assistant can use. Tick an ad set, then say "put the new ad in there", and it knows which one you mean.

### Audience estimate

A gauge showing how many people a proposed audience reaches per month, and whether that is too narrow, too broad, or about right. Check it before launching: a badly sized audience is the most common reason a well-built campaign fails to deliver.

### Launch review

Before the copilot creates anything, this card summarizes what the copilot will build: the campaign, the ad sets, the ads. You press **Confirm** inside the card. That creates everything **paused**; activating is a separate step afterwards.

Every confirm, activate and cancel goes back into the conversation, so the record of what you created lives in the chat, not only in the card.

### Creative picker

A thumbnail grid of your creative library, filterable by type, with multi-select. Pick an image by looking at it instead of describing its filename.

### Onboarding plan

An interactive checklist of your setup and strategy steps, with a progress ring. Tick as many items as you like; they save together.

## In assistants without card support

Everything still works. The same tools reply with text, plus a link where one makes sense. Nothing is missing; the reply is less pleasant to read.

Adsap can also switch cards off centrally. If they stop appearing, the tools behind them keep working.

## A note on how the assistant behaves around cards

After a card appears, the assistant says one short sentence and stops. It does not repeat the table as text. That is deliberate: the card is the answer, and a paragraph that restates it adds noise.

If you acted inside a card and want the assistant to pick up from there, ask. It reads what you did instead of asking you to type it again.