Worcadian MCP Service
Exposes three tools from the Worcadian JSON-RPC game server API (check_wordlist, gameday, results) as an MCP server using FastAPI, designed for deployment on Vercel.
README
Worcadian MCP Service
An MCP server, built with FastAPI, that wraps the
Worcadian JSON-RPC game server API (https://worcadian.vercel.app/rpc,
spec) as a set of MCP
tools. Designed to be deployed as a Python serverless function on Vercel.
What it exposes
The service speaks MCP over Streamable HTTP at /mcp, with three tools:
| MCP tool | Upstream JSON-RPC method | Arguments | Returns |
|---|---|---|---|
check_wordlist |
check |
words: list[str] |
{"WORD": true/false, ...} |
gameday |
gameday.current |
none | {"min_day": int, "max_day": int} |
results |
board.results |
game_day: int |
{"num_submissions", "best_score", "submissions": [...]} |
A plain GET /healthz route is also available for uptime checks.
How it works
app/worcadian_client.py— a minimal async JSON-RPC client (httpx) forhttps://worcadian.vercel.app/rpc. Upstreamerrorresponses are raised asWorcadianRPCError.app/mcp_server.py— aFastMCPinstance with the three tools above, each a thin wrapper around the client. Runs instateless_httpmode, since Vercel functions don't persist MCP session state between invocations, and with DNS rebinding protection explicitly disabled (that protection defaults to allowing onlylocalhost/127.0.0.1Host headers, which would reject every request once deployed on a real domain).app/server.py— the FastAPI app. It mounts the MCP app (which already serves/mcpby default) at the root, and wires itslifespanto start/stop the MCP session manager (required even in stateless mode, and not handled automatically by Starlette'sMount).api/index.py— the Vercel entrypoint; re-exports the FastAPIapp.
Running locally
cd worcadian-mcp
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements-dev.txt
uvicorn app.server:app --reload
The MCP endpoint is then available at http://127.0.0.1:8000/mcp. Point an MCP client
(Claude, Cursor, etc.) at that URL using its HTTP/Streamable-HTTP MCP server transport.
Testing
pip install -r requirements-dev.txt
pytest
Unit tests in tests/test_worcadian_client.py and tests/test_mcp_tools.py mock the
upstream HTTP call (via respx), so they run offline.
tests/manual_smoke_test.py is not part of the automated suite — it makes real calls to
the live Worcadian API (and optionally a deployed instance of this service) for a quick
human sanity check:
python tests/manual_smoke_test.py
# or, against a deployed MCP server:
MCP_BASE_URL=https://your-deployment.vercel.app python tests/manual_smoke_test.py
Deploying to Vercel
cd worcadian-mcp
vercel deploy
vercel.json routes all requests to api/index.py via the Python ASGI runtime. No
environment variables are required — the upstream API has no authentication.
After deploying, the MCP endpoint is https://<your-project>.vercel.app/mcp.
Out of scope
Only the three methods above are wrapped. The upstream API documents several other
methods (board.submit, analyse, seedwords.check, checkin.*, the 14 Numbers game,
etc.) that are not exposed by this service.
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。