@privacytracker/mcp
MCP server that exposes PrivacyTracker's app privacy data, changelogs, policy diffs, and sync status through read-only tools and resources for MCP-compatible clients.
README
@privacytracker/mcp
Model Context Protocol server for PrivacyTracker. Exposes the local privacy library — apps, privacy labels, change history, AI policy summaries — to any MCP-compatible client (Claude Desktop, Claude Code, Cline, etc.).
This is a read-only proxy. It talks to a running PrivacyTracker instance over its existing HTTP API; it does not touch the SQLite database directly and never mutates state.
Prerequisites
- A running PrivacyTracker instance (local
pnpm dev, Docker, or remote install). - Node.js 20 or newer on the machine running the MCP client.
Install
You don't need to clone this repo to use the server — every MCP client can spawn it via npx:
npx -y @privacytracker/mcp
For local development from a clone:
pnpm install
pnpm build
node dist/index.js
Configuration
The server reads three environment variables:
| Variable | Required | Default | Notes |
|---|---|---|---|
PRIVACYTRACKER_BASE_URL |
no | http://localhost:3000 |
Origin of the running PrivacyTracker instance. No trailing slash. |
PRIVACYTRACKER_ADMIN_TOKEN |
conditional | — | Required when the upstream server has AUDITOR_ADMIN_TOKEN set, OR when the MCP server is reaching it across a non-local hostname. Sent as x-auditor-admin-token on every request. |
PRIVACYTRACKER_TIMEOUT_MS |
no | 30000 |
Per-request timeout. |
A 401 response from the server triggers a clear MCP error explaining which side of the token contract is wrong.
Claude Desktop config
Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"privacytracker": {
"command": "npx",
"args": ["-y", "@privacytracker/mcp"],
"env": {
"PRIVACYTRACKER_BASE_URL": "http://localhost:3000",
"PRIVACYTRACKER_ADMIN_TOKEN": "..."
}
}
}
}
Restart Claude Desktop and the privacytracker server will appear in the tools menu.
Available tools
All tools are read-only and map 1:1 to existing PrivacyTracker API endpoints.
| Tool | Description |
|---|---|
list_apps |
Every tracked app with identity + change counts. Optional grouped view. |
get_app |
Full privacy dossier for one app (privacy types, categories, last snapshot, policy summary). |
search_apps |
iTunes Store search by name and/or bundle id. Returns App Store URLs. |
get_changelog |
Universal cross-app change feed with filtering (app, time, type, category) and pagination. |
get_app_changelog |
Full changelog for one app in chronological order. |
get_history_stats |
Per-quarter category trend and change-event sparkline for one app (Q1 2021 forward). |
get_policy_status |
Privacy-policy AI run status for one app (idle/running/done/error + log). |
get_policy_version_diff |
Line+word diff between a policy version and its predecessor. |
compare_apps |
Side-by-side privacy comparison of two apps (tracked or untracked URL). |
get_notifications |
30 most recent privacy-change notifications + unread count. |
get_sync_status |
Background sync scheduler state (running, last/next run, interval). |
Available resources
| URI | Description |
|---|---|
apps://list |
Flat JSON list of every tracked app. |
app://{appId} |
Full privacy dossier for one app. Replace {appId} with the numeric Apple track ID. |
Local development
pnpm install
pnpm dev # tsx watch mode
pnpm typecheck
pnpm test
pnpm lint
pnpm build
To smoke-test against a running PrivacyTracker instance:
PRIVACYTRACKER_BASE_URL=http://localhost:3000 pnpm start
The server speaks JSON-RPC over stdio. Use the MCP Inspector for interactive exploration:
npx @modelcontextprotocol/inspector node dist/index.js
Versioning
This package versions independently of PrivacyTracker itself. Tag a release (v0.x.y) on main and the publish workflow ships it to npm automatically.
License
Apache-2.0. See LICENSE.
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。