Building AI-native tools for martech & paid media strategy and operations.
An enterprise-grade AI operating system for paid media. Three autonomous agents handle data quality, multi-touch attribution, and budget execution. An MCP server connects Claude directly to live BigQuery data, agent outputs, and platform controls. A set of Claude Code skills handles strategy, planning, and ad-hoc analysis — all inside your own GCP environment.
Built to be flexible from the start. It works with local JSON files for simple setups and scales to BigQuery without changing how anything else works. Platform adapters (Meta, Google, LinkedIn, TikTok) are pluggable — use what you need, skip what you don't. The schema decisions, privacy constraints, budget guardrails, and human-approval workflows are all shaped by what actually breaks in production.
graph TD
human["👤 You + Claude"]
mcp["paid-media-mcp<br/>MCP server · TypeScript<br/>47 tools · 12 resources"]
skills["skills<br/>Claude Code slash commands<br/>/paid-media/*"]
watchdog["Watchdog<br/>data quality · signal capture"]
analyst["Analyst<br/>attribution · enrichment · insights"]
operator["Operator<br/>budget · audiences · suppressions"]
schema["paid-media-schema<br/>BigQuery · 7 schema layers<br/>reporting views · account analytics"]
platforms["Platform APIs<br/>Meta · Google Ads · LinkedIn · TikTok"]
human --> mcp & skills
mcp --> watchdog & analyst & operator
watchdog & analyst & operator --> schema
operator --> platforms
platforms -.->|daily exports| schema
| Repo | Role |
|---|---|
| paid-media-suite | Hub — AGENT.md system definition, SETUP.md, cross-repo coordination |
| paid-media-schema | BigQuery DDL — identity, platform, attribution, reporting views, account analytics |
| paid-media-mcp | MCP server — connects Claude to live data, agent outputs, and platform controls |
| paid-media-agent | Watchdog · Analyst · Operator — Python agents with Anthropic SDK tool-use loop |
| skills | Claude Code slash commands for strategy, measurement, creative, and campaign execution |