See your accounts
List every connected Google Ads account and get headline performance — spend, conversions, ROAS, CPA — across all of them, or 12-week trends and month-over-month comparisons for any one.
What AdTAO MCP does.
AdTAO MCP is the agent surface for a Google Ads decision engine. Point any MCP-capable AI host (Claude Desktop, Claude Code, Cursor, and more) at mcp.adtao.io, and your agent can read your accounts, diagnose problems, and — with your approval — execute fixes and measure what they did.
It is not a thin wrapper over the Google Ads API. The API gives you raw data; AdTAO gives you a grounded answer and a safe way to act on it.
Pattern intelligence
Archetype intelligence
AdTAO scans your account against 253 detection patterns maintained across the AdTAO network. A pattern might flag wasted spend — but it might also describe a healthy state, a transition in progress, or a prerequisite (e.g. fix tracking before touching bids). Every hit carries severity and, where we can quantify it, dollar impact — plus a recommended next step grounded in how similar accounts resolved it.
| Category | Example patterns |
|---|---|
| Measurement & trust | Conversion tracking drift after tag changes; goals that no longer match what's firing |
| Structure & scale | Manager/child hierarchy misalignment; campaigns that should be split or merged |
| Bidding & learning | Smart Bidding still in learning; tCPA/tROAS set against an impossible volume floor |
| Spend & efficiency | Budget pacing vs actual delivery; spend concentrated in low-intent segments |
| Competitive & query shape | Query themes bleeding spend; placement categories with network-known poor ROI |
Patterns aren't only point-in-time. AdTAO tracks how accounts move between observable states over weeks and quarters, so recommendations respect whether you're stabilizing, degrading, or recovering — not just what last week's CSV showed.
Outcome history
Predict before you act
When you stage a change, AdTAO draws on real intervention outcomes from accounts in similar situations — not model guesswork. If we don't have enough history for a confident answer, the tools return insufficient_data rather than a number.
Aside
Already using a 50-tool MCP?
GAQL wrappers turn your agent into a faster report writer. AdTAO is the grounded decision layer underneath — peer context, pattern diagnosis, and predict-before-you-act — that no wrapper can synthesize from one account's API key.
Section 01
What your agent can do
Ask in plain English
Ask any question about an account ("how is my CPA trending?", "which campaigns are wasting budget?") and get an answer grounded in real account data, not a guess.
Diagnose problems automatically
Scan an account against 253 detection archetypes — scored issues with severity, dollar impact, and a recommended fix. Surfaces the problems a human would take hours to find, in seconds.
Monitor budgets
Real-time budget pacing across the portfolio — who's over, who's under, who's on track — so nothing silently overspends or underspends.
Review and approve — you stay in control
Every change is staged for your approval before anything happens. Review the pending queue, approve or reject one at a time or in bulk, and the approved action executes. The agent never acts without sign-off.
Act at scale
Where your accounts sit under a shared manager hierarchy, an approved exclusion can be applied once at the parent level and inherited by every child account — one approval, enforced everywhere.
Competitive and keyword intelligence
See what competitors are running, extract their ad copy, identify top competing advertisers, and find keyword gaps and opportunities driven by a cross-account taxonomy.
Measure what happened
After a change executes, AdTAO measures the outcome — so next week's read includes "what last week's changes actually did."
Section 02
The primitive surface
The MCP exposes 16 core primitives, grouped by what they do. Tool names are stable; deprecations are append-only in the changelog.
Read · 9
gather and reason about account state
-
benchmark -
get_account_intelligence -
get_recent_changes -
get_intervention_outcome -
query_manager -
simulate_action -
briefing -
narrate -
get_approval_queue
Write · 4
stage and execute changes, always behind approval
-
stage_manager_action -
approve_and_execute -
amend_staged_action -
stage_goal_change
Meta · 3
session and feedback
-
session_state -
report_agent_friction -
subscribe_to_sync_complete
Section 03
What it deliberately does not do
We document our limits as plainly as our features.
It won't act without your approval.
Every write is staged first. This is the design, not a setting.
It won't claim a capability the Google Ads API doesn't have.
For example, Performance Max campaigns can't have individual channels (Display, YouTube, Discover) excluded via the API — so AdTAO won't pretend to. It surfaces the real levers (account- and parent-level placement, URL, and brand exclusions) and labels what's API-limited.
It won't invent data.
When an account hasn't finished syncing or there isn't enough data to be confident, it says so rather than guessing.
Continue exploring