Overview

This MCP (Model Context Protocol) server exposes Google Ads account data and campaign controls as tools that can be invoked programmatically by model-driven workflows to manage Facebook Ads. In practice it acts as a bridge: the server ingests Google Ads signals (audiences, performance metrics, budgets) and exposes a consistent, authenticated API surface compatible with MCP-enabled agents so models can read data and invoke actions to control programmatic Facebook Ads activity.

The server is useful when you want to automate cross-platform advertising workflows, for example using historical Google Ads performance to seed audience creation, pacing logic, or creative selection on Facebook. By wrapping credentials, rate-limits, auditing, and data transformations behind tool endpoints, it lets models and orchestration systems safely access advertising primitives without embedding service credentials inside prompts or application logic.

Features

Exposes Google Ads data as MCP-compatible tools for model-driven agents
Auth flow and credential management for Google Ads and Facebook APIs
Tool endpoints for campaign creation, budget updates, audience syncs, and reporting
Rate limiting, retry logic, and basic error normalization
Logging and audit trails for tool invocations
Local development via Docker and environment-driven configuration
Webhook and polling options for near-real-time sync between platforms

Installation / Configuration

Clone the repository and install dependencies (Node.js example):

git clone https://github.com/gomarble-ai/google-ads-mcp-server.git
cd google-ads-mcp-server
npm install

Environment configuration (example .env):

# Google Ads
GOOGLE_ADS_CLIENT_ID=your-google-client-id
GOOGLE_ADS_CLIENT_SECRET=your-google-client-secret
GOOGLE_ADS_DEVELOPER_TOKEN=your-developer-token
GOOGLE_ADS_REFRESH_TOKEN=your-refresh-token

# Facebook / Meta
FACEBOOK_ACCESS_TOKEN=your-facebook-access-token
FACEBOOK_AD_ACCOUNT_ID=act_123456789

# Server
MCP_SERVER_PORT=8080
LOG_LEVEL=info
RATE_LIMIT_PER_MINUTE=60

Start the server locally:

npm run build   # if using TypeScript
npm start

Or run with Docker:

# docker-compose.yml
version: "3.8"
services:
  mcp-server:
    image: gomarble/google-ads-mcp-server:latest
    ports:
      - "8080:8080"
    environment:
      - GOOGLE_ADS_CLIENT_ID=${GOOGLE_ADS_CLIENT_ID}
      - GOOGLE_ADS_CLIENT_SECRET=${GOOGLE_ADS_CLIENT_SECRET}
      - FACEBOOK_ACCESS_TOKEN=${FACEBOOK_ACCESS_TOKEN}

docker-compose up

Available Tools / Resources

The server exposes a set of tools (HTTP endpoints) that an MCP client or model can invoke. Typical tools include:

Tool name	Method	Path	Description
get_campaigns	GET	/tools/get_campaigns	List Google Ads campaigns and metrics
sync_audience	POST	/tools/sync_audience	Convert a Google Ads audience to Facebook Custom Audience
create_fb_campaign	POST	/tools/create_fb_campaign	Create or update a Facebook campaign from a spec
update_budget	POST	/tools/update_budget	Adjust budget pacing on Facebook based on rules
get_report	POST	/tools/get_report	Fetch aggregated performance metrics and anomalies
simulate	POST	/tools/simulate	Run a dry-run of actions and return expected outcomes

Example: request a campaign creation tool via curl

curl -X POST http://localhost:8080/tools/create_fb_campaign \
  -H "Content-Type: application/json" \
  -d '{
    "source_campaign_id":"123456",
    "name":"Lookalike_Campaign_From_GA",
    "budget":{"amount":500,"currency":"USD"},
    "targeting":{"countries":["US"],"audience_source":"google_audience_42"},
    "dry_run":false
  }'

Responses follow a normalized schema with status, request id, and standardized error codes to make them easy for models to consume.

Use Cases

Audience Syncing: Export a high-performing Google Ads remarketing list and create a matching Facebook Custom Audience to target with similar creatives. Example: schedule a nightly sync of “last 30-day converters” into Facebook using /tools/sync_audience.
Budget Rebalancing: Use Google Ads pacing signals (e.g., CPA spikes) to reduce spend on Facebook campaigns that are likely to underperform. Example: model reads /tools/get_report then calls /tools/update_budget to shift daily caps.
Cross-Platform A/B Testing: Generate variants from Google Ads top-performing creatives and programmatically create corresponding Facebook ad sets using /tools/create_fb_campaign, then fetch results via /tools/get_report for comparative analysis.
Automated Alerts & Remediation: Detect anomaly in CPC/CTR via /tools/get_report and trigger /tools/simulate to preview corrective actions, then apply them if simulated outcome meets thresholds.

Notes and Best Practices

Always run in dry-run mode when testing tool invocations to prevent accidental spend.
Store sensitive credentials in a secure secrets manager; the server accepts environment variables but can be configured to use vaults.
Respect ad platform policies and rate limits; the server includes basic throttling but you should tune it for production workloads.
Review audit logs before enabling autonomous model-driven actions in production.

For more details, sample workflows, and advanced configuration options, see the repository at https://github.com/gomarble-ai/google-ads-mcp-server.

Google Ads MCP Server for Programmatic Facebook Ads