adops-mcp
Cross-platform ad management MCP server for Google Ads and Meta Ads. Campaign analytics, A/B testing with z-test, anomaly detection, and budget reallocation. 15 tools, 60 tests.
README
AdOps MCP
AI-powered cross-platform ad management for the Model Context Protocol
Manage Google Ads and Meta Ads campaigns, analyze cross-platform performance, optimize budgets, and detect anomalies — all through AI assistants like Claude, Cursor, and VS Code.
Features
- Unified campaign management for Google Ads and Meta Ads
- 14 MCP tools covering the full advertising lifecycle
- 4 MCP resources for quick dashboard access
- Cross-platform performance reporting with normalized metrics
- AI-powered budget optimization with actionable recommendations
- Statistical anomaly detection (CPC spikes, CTR drops, spend surges)
- A/B test analysis with confidence scoring
- Industry benchmark comparison (9 verticals)
- Spend and conversion forecasting (7/14/30 days)
- Platform-specific creative specs reference
- 42 automated tests (unit + E2E user workflows), TypeScript strict mode, Zod validation
Quick Start
Install from npm
npm i adops-mcp-server
Add to your MCP client
{
"mcpServers": {
"adops": {
"command": "node",
"args": ["path/to/node_modules/adops-mcp-server/dist/index.js"]
}
}
}
Build from source
git clone https://github.com/enzoemir1/adops-mcp.git
cd adops-mcp
npm ci && npm run build
Tools
| Tool | Description |
|---|---|
platform_connect |
Register a Google Ads or Meta Ads account connection |
campaign_list |
List and filter campaigns across all connected platforms |
campaign_create |
Create a new campaign with unified parameters |
campaign_update |
Update campaign settings (budget, status, bidding, schedule) |
campaign_pause_resume |
Batch pause or resume up to 50 campaigns at once |
ads_report |
Generate unified cross-platform performance report |
budget_analyze |
Analyze budget allocation with optimization recommendations |
budget_reallocate |
Transfer budget between campaigns across platforms |
audience_insights |
Get demographic, geographic, and device breakdowns |
creative_specs |
Get platform-specific image, video, and text requirements |
anomaly_detect |
Detect performance anomalies with configurable sensitivity |
ab_test_analyze |
Compare two campaigns with statistical significance testing |
competitor_benchmark |
Compare your metrics against industry averages |
forecast_spend |
Forecast spend, conversions, and ROAS for the next period |
Resources
| Resource | Description |
|---|---|
ads://overview |
Cross-platform dashboard summary |
ads://campaigns |
All active campaigns with key metrics |
ads://budget |
Budget allocation across platforms |
ads://alerts |
Recent performance anomalies and warnings |
Configuration
All integrations are optional. The server works without API keys using local storage for campaign management and analytics.
| Variable | Platform | Required | Description |
|---|---|---|---|
GOOGLE_ADS_DEVELOPER_TOKEN |
For live sync | API developer token | |
GOOGLE_ADS_CLIENT_ID |
For live sync | OAuth2 client ID | |
GOOGLE_ADS_CLIENT_SECRET |
For live sync | OAuth2 client secret | |
GOOGLE_ADS_REFRESH_TOKEN |
For live sync | OAuth2 refresh token | |
GOOGLE_ADS_CUSTOMER_ID |
For live sync | Ad account ID (10 digits) | |
META_APP_ID |
Meta | For live sync | Facebook App ID |
META_APP_SECRET |
Meta | For live sync | Facebook App Secret |
META_ACCESS_TOKEN |
Meta | For live sync | System user access token |
META_AD_ACCOUNT_ID |
Meta | For live sync | Ad account ID (act_XXX) |
See .env.example for a complete template.
Unified Metrics
AdOps normalizes metrics across platforms into a single schema:
| Metric | Formula | Description |
|---|---|---|
| CTR | clicks / impressions x 100 | Click-through rate (%) |
| CPC | spend / clicks | Cost per click |
| CPM | spend / impressions x 1000 | Cost per 1000 impressions |
| ROAS | revenue / spend | Return on ad spend |
| CPA | spend / conversions | Cost per acquisition |
| Conversion Rate | conversions / clicks x 100 | Conversion rate (%) |
Platform field mapping:
| AdOps Field | Google Ads | Meta Ads |
|---|---|---|
spend |
cost_micros / 1,000,000 |
amount_spent |
impressions |
impressions |
impressions |
clicks |
clicks |
clicks |
conversions |
conversions |
actions[type=purchase] |
| Campaign > Ad Group | Ad Group | Ad Set |
Pricing
| Tier | Price | Features |
|---|---|---|
| Free | $0 | 1 platform, read-only reports, creative specs, 50 calls/day |
| Pro | $29/mo | 2 platforms, full CRUD, anomaly detection, budget optimization |
| Agency | $59/mo | Unlimited, forecasting, benchmarking, priority support |
Available on MCPize.
Development
npm run dev # Hot reload development
npm run build # Production build
npm test # Run 42 tests (unit + E2E)
npm run inspect # Open MCP Inspector
Testing
42 tests across 5 test suites:
- Storage: Connection CRUD, campaign search, metrics aggregation, batch inserts
- Analytics: Metric calculations, performance reports, forecasting, benchmarks
- Optimizer: Budget analysis, reallocation, scaling/pausing recommendations
- Anomaly: CPC spike detection, conversion drops, sensitivity levels, severity sorting
- E2E Workflow: 14 real user scenarios — connect platforms, create campaigns, generate reports, optimize budgets, run A/B tests, detect anomalies, forecast spend, benchmark vs industry
npm test
License
MIT License. See LICENSE for details.
Built by Automatia BCN.
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。