Google Ads
$18,420 spend | 4.1x ROAS
USD | Aug 1-31 | account attribution | refreshed 2m ago
Connect Adstudio through OAuth, discover the accounts available to you, and turn a plain question into reads across Google Ads, Meta Ads and TikTok Ads, with the evidence attached.
Question
How did my selected Google Ads and Meta Ads accounts perform this month?
I will make separate bounded reads and name each provider, account, period, currency, attribution, freshness, truncation, and unresolved coverage.
Google Ads
$18,420 spend | 4.1x ROAS
USD | Aug 1-31 | account attribution | refreshed 2m ago
Meta Ads
EUR 9,840 spend | 3.6x ROAS
EUR | Aug 1-31 | 7-day click | refreshed 3m ago
The accounts can be discussed together, but their measurement contexts stay separate.
Pick a platform
A conversation, backed by the account you named
In ChatGPT, add the public Adstudio MCP endpoint, complete OAuth, and start by discovering your active workspace integrations. Choose the workspace, platform, and account explicitly before asking for a scoped read.
ChatGPT menu labels vary by plan and rollout. You choose in Adstudio Settings → MCP whether ChatGPT can only read, or read and write, and it never reaches more than your own role allows.
Grounded work, less context switching
01
Start with a plain-language performance, entity, or creative question instead of building an export before the investigation begins.
02
ChatGPT can discuss separate provider reads together while account, currency, period, attribution, freshness, truncation, and coverage stay visible.
03
When an account is not explicit, the public MCP asks for the boundary rather than selecting one on your behalf.
Connection, with boundaries
ChatGPT receives a useful answer only after the source account is clear.
In Adstudio Settings → MCP, create your personal connection.
Pick Read only or Read and write. Your Adstudio role stays the limit.
Discover accounts, then explicitly name the workspace, platform, and account for a scoped read.
FAQ
Adstudio MCP
Connect once, choose the account you mean, and keep the source context attached to every result.