Can the AI agent pause or launch campaigns?
Yes, with a hard confirmation gate. The Claude agent can pause, scale, or launch campaigns through 5 destructive tools, but each tool call requires an explicit `confirmed: true` from you in the chat before it executes. Read-only tools run without confirmation.
product
Admaxxer is a DTC analytics platform with built-in Meta + Google ad ops. A question every operator asks before giving an AI access to their ad accounts: can the AI agent pause or launch campaigns? Yes — the Claude agent has write access to Meta and Google — but every destructive action is double-gated behind an explicit in-chat confirmation, so nothing fires without you seeing exactly what it will change.
## The short answer
Yes, with a hard confirmation gate. The Claude agent can pause, scale, or launch campaigns through 5 destructive tools, but each tool call requires an explicit `confirmed: true` from you in the chat before it executes. Read-only tools run without confirmation.
## Full answer
### The tool split
The agent has two categories of tools:
- **Read tools** (run freely): `list_campaigns`, `get_campaign_insights`, `get_account_insights`, `query_metrics`. These pull data from Meta, Google, and the Tinybird analytics pipes so the agent can answer questions about MER, ROAS, CAPI match rate, cohorts, forecasts, and more.
- **Destructive tools** (gated): `update_campaign`, `pause_all_low_roas`, and the campaign-creation path. These change ad account state on Meta or Google and require explicit confirmation.
### How the confirmation gate works
When you ask the agent to do something destructive ("pause all campaigns under 1.5 ROAS"), it plans the action and presents a structured preview: the exact campaigns, the intended change, and the reason. You respond with a confirm action that sets `confirmed: true` on the tool call. Only then does the agent execute. There is no "run-everything-now" mode that bypasses this.
### What it can do once confirmed
- Pause individual campaigns, ad sets, or ads
- Adjust daily or lifetime budgets
- Bulk-pause campaigns below a ROAS threshold
- Launch new campaigns based on a brief (targeting, creative, budget, objective)
### What it cannot do
The agent cannot change billing, add team members, rotate tokens, or delete historical data. It also cannot reach outside the ad-account scopes you authorized — it uses your stored OAuth refresh token or pasted token, so Admaxxer's permissions ceiling is whatever you granted on the Meta/Google side.
### Audit trail
Every destructive tool call is logged with the session ID, input arguments, your confirmation, and the API response. You can review the full action log on the agent's session page.
## Related questions
- [How does Admaxxer connect to Meta Ads?](/faq/how-does-admaxxer-connect-to-meta-ads)
- [How does Admaxxer connect to Google Ads?](/faq/how-does-admaxxer-connect-to-google-ads)
- [How does Admaxxer store my credentials?](/faq/how-does-admaxxer-store-my-credentials)
Related Questions
Does the Admaxxer AI agent make changes automatically?
No. Destructive actions like pausing, scaling, or launching campaigns require an explicit confirmed: true from you in the chat before the agent executes the tool call.
Which ad platforms can the Admaxxer agent write to?
The agent can write to any ad platform you have connected — Meta Ads and Google Ads in v1, with TikTok on the roadmap. Writes always use the OAuth scopes you already granted.
Can I audit what the AI agent did?
Yes. Every tool call, including confirmations and API responses, is logged to the chat session so you have a full audit trail of what the agent read or changed.
Still have questions?
See the full documentation or start a 7-day free trial — no credit card required.
Try Admaxxer Free