Skip to content
Bryan Silva
03ai-media-opsAll case studies

AI for data · MCP · Full-stack

AI-powered paid media operations: platform and MCP agent

A platform and an agent so the team sees every account in one place, produces standard reports and gets alerts.

Status
In use by the traffic team
Role
Design, development and deployment
Stack
  • React
  • TypeScript
  • Node.js
  • MySQL
  • Claude
  • MCP
  • Google Ads API
  • GA4
  • Meta Ads
  • TikTok Ads
  • LinkedIn Ads
  • n8n
  • WhatsApp
  • PM2
  • Nginx

Results

freed up/month
50+ h
platforms in one table
5 → 1
to set up a manager
~15 min

Problem

The team manages many accounts on Google, Meta, TikTok and LinkedIn; checking each one, catching problems in time and building consistent reports takes up much of the week.

Solution

Three layers: a metrics platform (React, Node.js, MySQL) with every account in one normalized table; n8n monitoring that only alerts when something needs action; and an agent on Claude Code connected via MCP to Google Ads, GA4, Tag Manager, Merchant Center, Meta and LinkedIn.

How it works

  1. 01Trigger

    Media APIs

    Google Ads, GA4, Meta, TikTok and LinkedIn, connected via OAuth.

  2. 02Process

    Daily sync

    Automatic sync with no duplicate data.

  3. 03Process

    Normalized table

    Single source of truth: 5 platforms in one MySQL table.

  4. 04Output

    Platform

    8 modules: overview, campaigns, PDF/PPT reports, AI insights, alerts and more.

  5. 05Process

    Monitoring

    n8n alerts on WhatsApp only when something needs action.

  6. 06AI

    MCP agent

    Claude Code connected to the accounts; read-only by default.

Inside the system

ai-media-ops/overview
Metrics platform: agency overview with portfolio health and consolidated performance (demo data)

Platform: overview (demo data)

ai-media-ops/report
Metrics platform: report builder with configurable metrics and sections (demo data)

Platform: report builder

Technical decisions

  • Single source of truth. Every platform in one normalized table, synced daily.
  • Read-only by default. Any change to an account needs explicit approval.
  • Reports with fixed rules. Same conversion source, same names, same format.
  • Ads × GA4 check. Missing data never becomes “zero results”.
  • Cost-controlled AI. Haiku for streaming chat, Sonnet for analyses, prompt caching and an AI budget per workspace.
  • Security. AES-256 credentials and multi-workspace with roles.
  • Agent with subagents. Commands for onboarding, standard reports, hypothesis-driven analysis, funnel strategy and transcription.

Result

A platform and an agent with access to every account, one reporting standard for the whole team and automatic alerts when a campaign needs attention.