agentic-ads-e2e

Agent skills and infrastructure for running Google Ads campaigns end-to-end, from creative generation to campaign reporting.

• .agents/skills/ — agent skills for image generation, video editing, and Google Ads (campaign summaries, Performance Max asset groups, Search campaigns and responsive search ads, experiments)
• ads-agent/ — ADK agent that runs the skills on Gemini through Vertex AI (see its README)
• google-ads-e2e/ — Terraform to deploy the official Google Ads MCP server on Cloud Run (see its README)

MCP servers

The skills expect these servers in Antigravity. Register them with agy mcp add; the config lives in ~/.gemini/config/mcp_config.json.

GEAP (Vertex AI) — image generation and prediction. These use OAuth. Create an OAuth 2.0 Web application client in your Google Cloud project with:

• Authorized JavaScript origin: https://antigravity.google
• Authorized redirect URI: https://antigravity.google/oauth-callback

Then add the servers and attach the client to each entry in mcp_config.json (agy mcp add has no OAuth flag):

agy mcp add geap-generate https://aiplatform.googleapis.com/mcp/generate
agy mcp add geap-predict  https://aiplatform.googleapis.com/mcp/predict

"geap-generate": {
  "serverUrl": "https://aiplatform.googleapis.com/mcp/generate",
  "oauth": { "clientId": "<CLIENT_ID>.apps.googleusercontent.com", "clientSecret": "<CLIENT_SECRET>" }
}

GenMedia (Veo, Omni, avtool) — video generation and editing. Local stdio servers. One script installs the binaries, checks ffmpeg, creates the output bucket, and registers them:

bash .agents/skills/genmedia-video-editing/scripts/setup.sh [--project PROJECT_ID] [--bucket BUCKET]

Google Ads MCP. Deployed to Cloud Run by the Terraform in google-ads-e2e/; its output prints the exact agy mcp add google-ads-mcp <url>/mcp command. See that README.

Verify with agy mcp list.

License

MIT. See LICENSE.