Google Ad Manager MCP Server
Enables AI assistants to manage Google Ad Manager campaigns, line items, creatives, and advertisers through natural language, automating ad operations that normally require countless clicks through the UI.
README
Google Ad Manager MCP Server
Automate Google Ad Manager with AI. An MCP server that lets AI assistants like Claude, ChatGPT, Gemini, Cursor, and VS Code manage your ad campaigns, line items, creatives, and more through natural language.
<p align="center"> <strong>Built by <a href="https://matious.com">Matious</a></strong> — We build custom AI tools and MCP servers for businesses. </p>
Why This Exists
Managing Google Ad Manager is tedious. Creating campaigns, uploading creatives, and configuring line items involves countless clicks through a complex UI.
This MCP server changes that. Connect it to Claude and manage your entire ad operations through conversation:
- "Create a new campaign for Nike ending December 31st"
- "Upload all creatives from this folder and associate them with the Display line item"
- "Check which orders are currently delivering"
No more clicking. Just tell Claude what you need.
Features
- Order Management: List, create, and manage orders
- Line Item Management: Create, duplicate, and configure line items
- Creative Management: Upload images, associate with line items, bulk upload
- Advertiser Management: Find, create, and list advertisers
- Verification Tools: Verify line item setup, check delivery status
- Campaign Workflow: Complete campaign creation in one operation
Installation
From PyPI (Recommended)
pip install google-ad-manager-mcp
Or with uv:
uv pip install google-ad-manager-mcp
From Source
git clone https://github.com/MatiousCorp/google-ad-manager-mcp.git
cd google-ad-manager-mcp
pip install -e .
Dependencies
- FastMCP: MCP server framework with native middleware support
- googleads: Google Ad Manager SOAP API client
Configuration
The server uses environment variables for configuration:
| Variable | Description | Required |
|---|---|---|
GAM_CREDENTIALS_PATH |
Path to service account JSON | Yes |
GAM_NETWORK_CODE |
Ad Manager network code | Yes |
GAM_MCP_TRANSPORT |
Transport mode: stdio or http |
No (default: stdio) |
GAM_MCP_HOST |
Server host (HTTP mode only) | No (default: 0.0.0.0) |
GAM_MCP_PORT |
Server port (HTTP mode only) | No (default: 8000) |
GAM_MCP_AUTH_TOKEN |
Authentication token (HTTP mode only) | No (auto-generated if not set) |
Authentication
The server implements Bearer token authentication using FastMCP native middleware, following MCP security best practices.
Security Features
- FastMCP Native Middleware: Uses FastMCP 2.x middleware for proper MCP lifecycle management
- Cryptographically secure tokens: Generated using
secrets.token_hex(32) - Timing attack prevention: Uses constant-time comparison (
hmac.compare_digest) - Tool-level authentication: Auth validated on every tool call
- Audit logging: All authentication failures logged
How It Works
Authentication is enforced at the tool level using FastMCP's middleware system:
- When a tool is called, the middleware validates the
Authorizationheader - If no token is configured (
GAM_MCP_AUTH_TOKENnot set), requests are allowed - Invalid or missing tokens return a
ToolErrorwith a helpful message
Setup
For remote deployments, set a fixed authentication token:
# Generate a secure token
python -c "import secrets; print(secrets.token_hex(32))"
# Set it as environment variable
export GAM_MCP_AUTH_TOKEN="your-generated-token"
If not set, a random token is generated at startup and displayed in the logs.
Clients must include the token in the Authorization header:
Authorization: Bearer your-generated-token
Endpoints
| Endpoint | Description |
|---|---|
/mcp |
MCP protocol endpoint (auth validated on tool calls) |
Running the Server
Local Development
# Using the installed command
gam-mcp
# Or directly with Python
python -m gam_mcp.server
# With custom configuration
GAM_NETWORK_CODE=12345678 GAM_MCP_PORT=9000 gam-mcp
Docker Deployment
The Docker image runs as a non-root user (appuser) for security.
Build the Image
docker build -t google-ad-manager-mcp .
Run the Container
# Basic usage with credentials mounted
docker run -d \
--name gam-mcp \
-p 8000:8000 \
-v /path/to/your/credentials.json:/app/credentials.json:ro \
-e GAM_NETWORK_CODE=YOUR_NETWORK_CODE \
google-ad-manager-mcp
# With authentication token (recommended for production)
docker run -d \
--name gam-mcp \
-p 8000:8000 \
-v /path/to/your/credentials.json:/app/credentials.json:ro \
-e GAM_NETWORK_CODE=YOUR_NETWORK_CODE \
-e GAM_MCP_AUTH_TOKEN=$(python -c "import secrets; print(secrets.token_hex(32))") \
google-ad-manager-mcp
# With custom port
docker run -d \
--name gam-mcp \
-p 9000:8000 \
-v /path/to/your/credentials.json:/app/credentials.json:ro \
-e GAM_NETWORK_CODE=YOUR_NETWORK_CODE \
-e GAM_MCP_PORT=8000 \
google-ad-manager-mcp
View Logs
# View startup logs (includes generated auth token if not set)
docker logs gam-mcp
# Follow logs
docker logs -f gam-mcp
Docker Compose
Create a docker-compose.yml file:
version: '3.8'
services:
gam-mcp:
build: .
ports:
- "8000:8000"
volumes:
- ./credentials.json:/app/credentials.json:ro
environment:
- GAM_NETWORK_CODE=YOUR_NETWORK_CODE
- GAM_MCP_AUTH_TOKEN=your-secure-token
restart: unless-stopped
Run with:
docker-compose up -d
Verify the Container
# Check container is running
docker ps
# Test the endpoint
curl -X POST http://localhost:8000/mcp \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-d '{"jsonrpc": "2.0", "method": "initialize", "params": {"protocolVersion": "2024-11-05", "capabilities": {}, "clientInfo": {"name": "test", "version": "1.0"}}, "id": 1}'
Cloud Deployment (Railway, Fly.io, etc.)
-
Set environment variables in your cloud provider:
GAM_CREDENTIALS_PATH: Path to credentials (or use secrets)GAM_NETWORK_CODE: Your Ad Manager network codeGAM_MCP_AUTH_TOKEN: A secure authentication token
-
Deploy using the included Dockerfile
Connecting to AI Assistants
Claude Desktop (uvx - Recommended)
The easiest way to use this server with Claude Desktop. Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"google-ad-manager": {
"command": "uvx",
"args": ["google-ad-manager-mcp"],
"env": {
"GAM_CREDENTIALS_PATH": "/path/to/your/credentials.json",
"GAM_NETWORK_CODE": "YOUR_NETWORK_CODE"
}
}
}
}
Claude Desktop (Docker)
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"google-ad-manager": {
"command": "docker",
"args": [
"run", "-i", "--rm",
"-e", "GAM_NETWORK_CODE",
"-v", "/path/to/credentials.json:/app/credentials.json:ro",
"google-ad-manager-mcp"
],
"env": {
"GAM_NETWORK_CODE": "YOUR_NETWORK_CODE"
}
}
}
}
Claude Desktop (HTTP Mode)
If running the server in HTTP mode:
{
"mcpServers": {
"google-ad-manager": {
"url": "http://localhost:8000/mcp"
}
}
}
Remote Server with Authentication
If deploying remotely with authentication enabled:
{
"mcpServers": {
"google-ad-manager": {
"url": "https://your-server.com/mcp",
"headers": {
"Authorization": "Bearer your-secure-token"
}
}
}
}
Other MCP Clients
This server works with any MCP-compatible client, including:
- ChatGPT Desktop - OpenAI adopted MCP in March 2025
- Cursor - AI-powered IDE with MCP support
- VS Code - Via MCP extensions
- Windsurf, Zed, Codeium - Various IDE integrations
Refer to each client's documentation for MCP server configuration.
Testing with MCP Inspector
# Without authentication
npx @modelcontextprotocol/inspector http://localhost:8000/mcp
# With authentication (set header in Inspector UI)
# Header: Authorization
# Value: Bearer your-token
Available Tools
Order Tools
| Tool | Description |
|---|---|
list_delivering_orders |
List all orders with delivering line items |
get_order |
Get order details by ID or name |
create_order |
Create a new order |
find_or_create_order |
Find existing or create new order |
Line Item Tools
| Tool | Description |
|---|---|
get_line_item |
Get line item details |
create_line_item |
Create a new line item |
duplicate_line_item |
Duplicate an existing line item |
update_line_item_name |
Rename a line item |
list_line_items_by_order |
List all line items for an order |
Creative Tools
| Tool | Description |
|---|---|
upload_creative |
Upload an image creative |
associate_creative_with_line_item |
Associate creative with line item |
upload_and_associate_creative |
Upload and associate in one step |
bulk_upload_creatives |
Upload all creatives from a folder |
get_creative |
Get creative details |
list_creatives_by_advertiser |
List creatives for an advertiser |
Advertiser Tools
| Tool | Description |
|---|---|
find_advertiser |
Find advertiser by name |
get_advertiser |
Get advertiser details |
list_advertisers |
List all advertisers |
create_advertiser |
Create a new advertiser |
find_or_create_advertiser |
Find or create advertiser |
Verification Tools
| Tool | Description |
|---|---|
verify_line_item_setup |
Verify line item configuration |
check_line_item_delivery_status |
Check delivery progress |
verify_order_setup |
Verify entire order setup |
Workflow Tools
| Tool | Description |
|---|---|
create_campaign |
Complete campaign creation workflow |
Example Usage with Claude
User: List all delivering orders
Claude: [Uses list_delivering_orders tool]
Here are the currently delivering orders:
1. Campaign IPhone 17 Pro 2025/2026 (ID: 123456)
- Display line item: 45,000 impressions delivered
User: Create a new campaign for "ACME Corp" ending December 31, 2025
Claude: [Uses create_campaign tool]
I'll create the campaign with:
- Advertiser: ACME Corp
- Order: ACME Campaign 2025
- Line Item: Display
- End Date: December 31, 2025
Campaign created successfully!
- Order ID: 789012
- Line Item ID: 345678
- 4 creatives uploaded and associated
Development
Setup
# Clone the repository
git clone https://github.com/MatiousCorp/google-ad-manager-mcp.git
cd google-ad-manager-mcp
# Install with dev dependencies
pip install -e ".[dev]"
Running Tests
# Run all tests
pytest
# Run with coverage
pytest --cov=gam_mcp --cov-report=html
# Run specific test file
pytest tests/test_utils.py
Code Quality
# Run linter
ruff check .
# Run linter with auto-fix
ruff check . --fix
Roadmap
The following features are planned for future releases:
Near-term
- [ ] Ad Unit Management - List, get, and create ad units with hierarchy support
- [ ] Placement Management - Manage inventory placements and targeting
- [ ] Line Item Status Control - Pause, resume, archive, and approve line items
- [ ] Forecast & Availability - Check inventory availability and forecast impressions
- [ ] Creative Preview Links - Generate preview URLs for creative-line item combinations
Medium-term
- [ ] Advanced Targeting - Geographic, device, daypart, and custom key-value targeting
- [ ] Reporting Tools - Generate and retrieve performance reports
- [ ] Bulk Operations - Batch updates for line items, creatives, and targeting
- [ ] HTML5/Video Creatives - Support for rich media and video creative uploads
Long-term
- [ ] Audience Management - Create and manage audience segments
- [ ] User & Permissions - Manage users, roles, and order assignments
- [ ] Yield Management - Configure yield groups and optimization
- [ ] Custom Reporting - Scheduled reports with export capabilities
Community Requests
Have a feature request? Open an issue to suggest new functionality.
Contributing
Contributions are welcome! Please see CONTRIBUTING.md for guidelines.
Changelog
See CHANGELOG.md for version history.
API Version
Uses Google Ad Manager SOAP API version v202411.
License
MIT - see LICENSE for details.
Need a Custom MCP Server?
This project is built and maintained by Matious.
We specialize in building custom AI tools and MCP servers that integrate with your existing systems. Whether you need to connect Claude to your CRM, ERP, ad platforms, or internal tools — we can help.
What we build:
- Custom MCP servers for any API or platform
- AI-powered automation workflows
- Claude integrations for business operations
Get in touch: matious.com
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。