golf-reports
Pulls golf data from Arccos, GHIN, and 18Birdies, and generates interactive HTML and PDF round reports with shot maps and stats via local MCP tools.
README
Golf Reports
A self-contained golf-data skill: pulls your own data from Arccos (shot-level
- strokes-gained), GHIN (official handicap + scores), and 18Birdies (export import), then renders any round to an HTML report (satellite shot map + stats) and a shot-map PDF. Login-based — no browser DevTools. Runs locally, so it works inside the Claude Desktop app (Chat / Cowork / Code) via a local MCP server.
Why local: the Cowork cloud sandbox can't reach the golf APIs or your files. The engine runs on your machine and exposes tools to Claude. (Architecture vetted with Codex.)
What's inside
ingest/ pull_arccos.py · pull_ghin.py · pull_18birdies.py (login-based, no DevTools)
render/ gen_combined · gen_stats · gen_satellite · gen_gps_pdf · build_round_pages
mcp/ server.py (MCP tools: list_rounds, round_stats, render_round, sync_*, …)
skills/ golf-reports/SKILL.md
setup.py one-time login wizard
manifest.json / .claude-plugin/plugin.json packaging
1. One-time setup (no DevTools)
pip install -r requirements.txt
python3 setup.py # prompts for Arccos + GHIN email/password, optional 18Birdies export
Credentials are stored only on your machine (~/.arccos_creds.json,
~/.ghin_creds.json, chmod 600) and never uploaded. Arccos stores a long-lived
accessKey (password discarded); GHIN stores the login (needed for its 12h token).
2. Use it in the Claude Desktop app
Cowork / Chat (recommended for non-devs): install as a one-click Desktop Extension (the local MCP server) — Customize → Browse plugins/extensions → add this package. Then ask: "sync my golf data", "how was my last round?", "render my latest round as a PDF I can send."
Manual MCP config (alternative): add to claude_desktop_config.json:
{ "mcpServers": { "golf-reports": {
"command": "python3", "args": ["/ABSOLUTE/PATH/golf-reports/mcp/server.py"],
"env": { "GOLF_STORE": "/ABSOLUTE/PATH/golf-data",
"GOLF_INGEST": "/ABSOLUTE/PATH/golf-reports/ingest",
"PYTHONPATH": "/ABSOLUTE/PATH/golf-reports/render" } } } }
Claude Code (Code tab / CLI): /plugin marketplace add <your-gh>/golf-reports
then /plugin install golf-reports, or copy skills/golf-reports into ~/.claude/skills/.
3. CLI (works anywhere, no Claude)
cd render
python3 build_round_pages.py <store-dir-or-git-url> ./out --all --split
# -> <course>_<date>_report.html + _shotmaps.pdf (+ _stats.html + _map.html)
Notes & provenance
- Strokes gained (by category / 15-yd band / hole) is measured (Arccos, vs
scratch). Putting make % and approach proximity are derived heuristics.
Peer carry is vs a modeled ~12-HCP table — edit
render/peer_config.json. - Satellite tiles (Esri) load only when the HTML is opened in a browser; the PDF is the reliable offline/shareable artifact. GPS is optional at every layer (map omitted / PDF falls back to a distance schematic).
- Privacy: credentials, raw data, and reports stay local. Nothing is published.
Verify on Desktop (do these first — the parts that can't be tested headless)
- Install the extension and confirm a tool (e.g.
list_rounds) appears + runs from the Cowork tab on your OS/plan. - From the running MCP server, confirm
sync_arccos/sync_ghinreach the APIs and write files. - On a clean machine without developer Python, confirm
matplotlibPDF generation works after install (bundle an embedded Python / PyInstaller if not — the riskiest packaging step).
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。