meta-ads-mcp-server
MCP Server for the Meta Marketing API. Gives Claude Desktop direct access to your ad account data — campaign performance, creative analysis, audience breakdowns, and budget pacing.
README
Meta Ads MCP Server
MCP Server for the Meta Marketing API. Gives Claude Desktop direct access to your ad account data — campaign performance, creative analysis, audience breakdowns, and budget pacing.
Tools
| Tool | Description |
|---|---|
get_campaigns |
List campaigns with status, objective, and budget info |
get_campaign_performance |
Core KPIs: spend, impressions, CTR, CPC, CPM, CPA, ROAS |
get_creative_performance |
Creative-level metrics including video hook rate and thruplay rate |
get_ad_creatives |
Inline-render ad creative images (full-resolution by default, image_size: "thumbnail" for 64×64 previews) |
get_audience_breakdown |
Performance broken down by age and/or gender |
get_budget_status |
Budget pacing and daily spend status for active campaigns |
get_ad_accounts |
List all ad accounts the token has access to |
get_adsets / get_ads |
List ad sets / ads with optional filtering |
debug_* |
Diagnostic helpers for troubleshooting field/breakdown issues |
Setup
1. Install & Build
git clone https://github.com/YOUR_USERNAME/meta-ads-mcp-server.git
cd meta-ads-mcp-server
npm install
npm run build
2. Configure Environment
cp .env.example .env
Edit .env with your credentials:
META_ACCESS_TOKEN=your_meta_api_token
META_AD_ACCOUNT_ID=act_123456789
Getting your credentials:
- Ad Account ID: Found in Meta Ads Manager URL or account dropdown. Prefix with
act_. - Access Token: Generate at Meta Graph API Explorer with
ads_readpermission. Short-lived tokens (~2h) can be exchanged for 60-day long-lived tokens viaGET /oauth/access_token?grant_type=fb_exchange_token&client_id={APP_ID}&client_secret={APP_SECRET}&fb_exchange_token={SHORT_TOKEN}.
The
.envfile takes precedence over environment variables passed by Claude Desktop. If Claude Desktop'sclaude_desktop_config.jsoncontains a staleMETA_ACCESS_TOKEN, the value in.envwins. Update.envto rotate the token without touching Claude Desktop config.
3. Add to Claude Desktop
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"meta-ads": {
"command": "node",
"args": ["/absolute/path/to/meta-ads-mcp-server/dist/index.js"]
}
}
}
On macOS with Homebrew Node, use
/opt/homebrew/bin/nodeas the command.
Restart Claude Desktop. The tools will be available automatically.
Usage Examples
Ask Claude things like:
- "Show me my active campaigns"
- "How did campaign 120210001 perform last 7 days?"
- "Compare creative performance for my Summer Sale campaign"
- "Break down my campaign audience by age and gender"
- "Are any campaigns overspending today?"
Architecture
src/
├── index.ts → MCP server setup, tool registration
├── types.ts → Shared interfaces (MetaInsight, MetaAction, etc.)
├── fields.ts → Reusable Meta API field selections
├── formatters.ts → KPI calculations, number formatting
├── meta-client.ts → Meta API client (loads .env with override:true)
└── tools/
├── campaigns.ts → get_campaigns
├── performance.ts → get_campaign_performance
├── creatives.ts → get_creative_performance
├── ad-creatives.ts → get_ad_creatives (inline image rendering)
├── audience.ts → get_audience_breakdown
├── budget.ts → get_budget_status
├── accounts.ts → get_ad_accounts
├── adsets.ts → get_adsets
├── ads.ts → get_ads
└── debug.ts → debug helpers
Meta API Notes
- Budgets are in cents.
daily_budget: "5000"= €50.00 - Actions are arrays. Always parsed with
extractAction(), never accessed directly - Breakdowns multiply rows.
age,gendergives ~14 rows per campaign (7 age groups × 2 genders) - Video fields may be missing. Non-video ads won't have
video_thruplay_watched_actions - date_preset vs time_range: Never send both — Meta throws an error
License
MIT
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。
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。
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。
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。
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。
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。
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。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。