The client ad report without the exports

A marketing agent that pulls live Google Ads and Meta data through the Supermetrics MCP server into Claude, fills the client's report in Google Sheets, flags pacing problems in Slack and pauses campaigns only after approval

The problem

A small agency in Bangkok or Singapore runs ads for a dozen clients across Google, Meta and TikTok, and every Monday someone exports each platform, pastes the numbers into the client's sheet and writes two lines of commentary. By the time an overspending campaign is noticed in that sheet, half the week's budget is gone.

How the agent works

1. It starts

   The agency connects the Supermetrics MCP server to Claude for the ad accounts it manages.

2. The agent works

   Each morning a scheduled run asks Claude for yesterday's spend, conversions and cost per result for every active campaign; Supermetrics returns normalised data with its source and timestamp, and Claude compares each campaign against the monthly budget and the client's target. Campaigns pacing ahead of budget or under target get a line in the agency's Slack channel with the reason Claude sees in the data. On Monday the same run writes the weekly numbers and a short commentary into each client's Google Sheet report. When the account manager agrees that a campaign should be paused or a bid adjusted, Claude proposes the change through Supermetrics and it waits on the approval page until a person approves it; new campaigns are always created paused.

3. The result

   Nothing changes in a live account without that approval.

Worked example

Anthropic's case study on Supermetrics describes the build: Claude calls Supermetrics tools through its MCP server, the server returns normalised data with source, timestamp and transformation logic, daily pacing checks and anomaly alerts can arrive in Slack, marketers can create and edit campaigns in Google Ads, Meta, Microsoft Advertising, TikTok, LinkedIn and Snapchat by conversation, new campaigns are created paused, customers set approval levels per account, and an approval page lets stakeholders approve or reject each change. The worked example is Layer, a 25 person agency: a report that pulled from five sources and took 10 hours of billable work is now done in 20 minutes, 17 of which go to dropping the finished data into the client's Google Sheet. Supermetrics reports 250% average month over month growth in active users since the Claude connector launched. Those are Supermetrics' and Layer's figures; your own report time is the one to measure.

Sources

• https://claude.com/customers/supermetrics