Now live for Google AdsHow it works

Google Ads for AI agents.

Adjar is the Google Ads control layer for AI agents. Agents work through the adjar CLI — propose changes, review the diff, apply — safely and reversibly, like the Google Ads console does for humans.

google-ads · Q2 prospecting
EMEA · Search · Brand
daily budget €120 €180
In-place
US · Performance Max · Catalog
target CPA $32 $28
In-place
APAC · Search · Generic
ad rotation optimize rotate evenly
Resets learning

This restarts the learning phase. Adjar flagged it for review before apply.

3 changes · 1 needs review
Roll back Approve
How it works

Propose. Review. Apply.

Three steps between an idea and a change that’s live on your account — and reversible.

Propose

Tell Adjar what you want in plain language, or let it surface opportunities from your performance data. Adjar turns that into a concrete set of changes.

  • Pause everything under a 2x ROAS
  • Scale the winners in EMEA
  • Test three new headlines on the top spenders
Use cases

What you’ll actually do with Adjar.

The everyday moves of paid search — scaled across your whole account, and reversible.

Scale what’s working

Found a structure that converts? Define it once and fan it out across products, regions, and audiences. Adjar replicates the winners and shifts budget toward them — without a hundred trips through the UI.

  • Clone a proven structure to new regions
  • Shift budget toward the winners
  • Stay inside safe, learning-preserving edits
fan out structure
winning-structure.base
EMEA42 campaigns
North America38 campaigns
APAC27 campaigns
LATAM19 campaigns
Config is the source of truth

Your whole ad account, as config your AI can drive.

Adjar represents your entire account — campaigns, ad groups, keywords, creatives, assets, and conversions — as declarative config. Keep it in one file or split it across many; a root file composes the rest. It's plain text, version-controlled, and structured so a language model can read and reason about the whole account. Point your coding agent at it and let it propose changes in plain language — Adjar turns the edits into a reviewable plan, flags anything that would reset learning, and applies it safely.

db-ads-config/
# account.toml — the root configplatform = "google"imports = [  "google/assets.toml",  "google/conversions.toml",  "google/campaigns/*.toml",][account]id = "7272244390"sitelinks = [ "152179909917" ]callouts = [ "78942378600", "111050946545" ]structured_snippets = []
your coding agent
~/ads $ agent "pause keywords under 2x ROAS in JIT access, then scale EMEA 20%"

 Read campaigns.toml
 Edited campaigns.toml
   3 keywords status ENABLED "PAUSED"
   1 campaign daily_budget_usd 200 → 240

Run `adjar plan` to review the diff before anything goes live.
One file, every agent. No SDK to learn.
Claude CodeCodexGemini CLICursorAny MCP client

Managing paid search shouldn't mean a thousand clicks.

Start with Google Ads

More platforms coming.