github-mcp
Provides four GitHub superpowers (repo-scorecard, compare-repos, commit-heatmap, trending-mcp) as MCP tools with React widgets for repository health analysis, comparison, commit activity visualization, and trending MCP servers.
README
github-mcp
Four GitHub "superpowers" as an mcp-use MCP App — each a tool + React widget that renders inside Claude, ChatGPT, or any MCP Apps-compatible client. Write once, render everywhere.
| Tool | Ask it | Widget |
|---|---|---|
repo-scorecard |
"how healthy is vercel/next.js?" | Health ring + stat grid + contributors |
compare-repos |
"next.js vs remix?" | Side-by-side table, winner per metric |
commit-heatmap |
"how active is langchain?" | 52-week commit heatmap |
trending-mcp |
"hottest MCP servers this week?" | Ranked list with star deltas |
Quick start
npm install
cp .env.example .env # then paste a GITHUB_TOKEN
npm run dev # server on :3000, Inspector at /inspector
Open http://localhost:3000/inspector, list tools, and call e.g. repo-scorecard with { "repo": "vercel/next.js" }.
GitHub token
A token is strongly recommended — it lifts the REST rate limit from 60/hr → 5,000/hr.
- A fine-grained PAT with Public Repositories (read-only) is enough, or a classic PAT.
- Set it as
GITHUB_TOKEN(in.envlocally, in project settings when deployed).
Project layout
github-mcp/
├── index.ts # MCPServer + all 5 tool definitions
├── src/
│ ├── github.ts # shared client: auth, TTL cache, 202 retry, pagination
│ └── metrics.ts # health score, bus factor, median issue-close
├── resources/ # widgets (auto-discovered)
│ ├── scorecard.tsx
│ ├── compare.tsx
│ ├── heatmap.tsx
│ └── trending.tsx
├── public/
└── .env # GITHUB_TOKEN=...
How it works
src/github.ts— one thin client every tool uses. HandlesBearerauth, a per-path TTL cache (light tools ~1–5 min,star-growth/trendingcached hard), the/stats/*202 Acceptedretry, andLink-header pagination.- Caching TTLs — scorecard/compare/heatmap: 1–5 min;
trending-mcp: 1 h (search has its own ~30/min limit). - Widgets — each reads typed props via
useWidget<Props>(), adapts to light/dark viauseWidgetTheme(), and wraps in<McpUseProvider autoSize>.
Build & deploy
npm run build # production build (bundles widgets + server)
npm run deploy # deploy via mcp-use
To deploy on Manufact: push this repo, connect it via the Manufact GitHub App, set GITHUB_TOKEN in project settings, and every push auto-deploys with a preview URL. Then publish the connector to Claude / the ChatGPT Apps Store — same codebase, no host-specific changes.
Windows note:
npm run buildbundles all widgets and completes the TS build successfully; the finaltsctype-check step can fail on Windows because mcp-use invokestscvia an unquoted path containingC:\Program Files. Runnpx tsc --noEmitdirectly to type-check.
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。