What was in the way
Google Ads API is slow and rate-limited. When you're analyzing campaigns with Claude or running optimization loops, you hit walls constantly. Needed a way to work at local speed without API friction.
Founder
An AI-operable mirror of a Google Ads account that never edits blind.
Pulls a Google Ads account into a local mirror an AI can actually analyse.
Win: Analysis, applier and change log are separate modules, so nothing edits an account blind.
The results
- Agentic ad ops
- An AI maintaining live ad spend
- Nothing edits blind
- Analysis, applier and log split
- 2 ad platforms
- Mirrored locally for AI analysis
What it is
A local mirror of a Google Ads account so an AI can read and reason about campaign data without fighting the Ads UI. It goes past Google: there is a Facebook client and a cross-platform layer alongside, plus a dashboard, a landing-page set and an intent scorer.
What I built
13 Python modules, and the design decision worth naming is the separation: analysis, an applier, and a change database are three different things, so a proposed change is inspected and logged before it touches a live account. That matters more than it sounds — an AI with write access to ad spend is the exact place where "it seemed right" gets expensive.
The outcome
Built fast and used on real accounts. The performance numbers it produced are recorded against the campaigns they came from rather than quoted here, because a CPA figure with no attached export is a story, not a result.