An AI monitoring system that watches a live Google Ads account every hour, detects campaign anomalies the moment they happen, and sends an instant Telegram alert with a plain-English explanation and a specific fix. On its first live run it caught a 540% CPC spike before any manual review.
See It In Action
The Problem
Here's a scenario every performance marketer has lived: you open Google Ads on Thursday and notice a campaign has been overspending since Monday. Four days of wasted budget. The client calls before you do.
The problem isn't that the data wasn't there — it was. The problem is nobody was watching at 2am on Tuesday when the CPC started climbing. This agent watches every hour, every day, while you sleep. When something breaks it sends you a message in plain English — not a dashboard to log into, not a report to run, just: here's what happened, here's why it matters, here's what to check first.
The Solution
An automated Python script connects to Google Ads via MCP and runs every hour via Windows Task Scheduler. The moment it detects an anomaly it fires a Telegram alert — no dashboards, no manual checks, no delays.
Connects to Google Ads via MCP every hour and pulls today's live campaign data — spend, impressions, clicks, CTR, CPC, conversions — then automatically calculates a 7-day average baseline for every metric across every active campaign
Runs 5 detection rules against the baseline — CPC spike above 30%, CTR drop below 25%, daily overspend above 120%, daily underspend below 60% after 3pm, zero conversions on a campaign that normally converts — flags every breach instantly
Sends each flagged anomaly to Claude API which writes a 3-line plain-English alert — what is happening, why it matters, and the first specific thing to check — then fires it as a Telegram message to your phone within seconds
The Impact
Project Visual
Tools & Stack