GoalCast MCP Server
Enables registering predictions and checking pool status for match predictions on the Injective testnet via MCP.
README
GoalCast MCP Server
A small, standards-compliant MCP server (using the official
@modelcontextprotocol/sdk, Streamable HTTP transport) that wraps the
PredictionPool CosmWasm contract deployed on Injective testnet. It exposes
exactly two tools:
| Tool | What it does |
|---|---|
register_prediction |
Signs and broadcasts RegisterPrediction to the contract |
get_pool_status |
Reads the current home/draw/away split for a match |
This is what GoalCast's Next.js app (lib/injective.ts) connects to as a
real MCP client — not a hand-rolled REST bridge.
Setup
cd mcp-server
npm install
cp .env.example .env
Fill .env:
INJECTIVE_PRIVATE_KEY=<testnet wallet private key, no 0x needed either way>
INJECTIVE_NETWORK=testnet
PREDICTION_POOL_CONTRACT_ADDRESS=inj1... # from DEPLOY.md step 4
PORT=8787
Use a dedicated testnet-only wallet here — this key signs every prediction the server submits. In this hackathon MVP the MCP server's wallet is the signer for all predictions (see "Known limitation" below).
Run it:
npm run dev
You should see:
GoalCast MCP server listening on http://localhost:8787/mcp
Connect the Next.js app to it
In the main app's .env.local:
INJECTIVE_MCP_SERVER_URL=http://localhost:8787/mcp
Restart npm run dev in the app root. lib/injective.ts will now call the
real tools instead of its simulated fallback.
Test it directly (without the app)
curl -X POST http://localhost:8787/mcp \
-H "Content-Type: application/json" \
-H "Accept: application/json, text/event-stream" \
-d '{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": { "name": "get_pool_status", "arguments": { "matchId": "m-usa-mex-01" } }
}'
Or point the MCP Inspector at
http://localhost:8787/mcp for a UI to browse and call the tools.
Known limitation (worth mentioning in your demo/judging notes)
Right now the MCP server's own wallet signs every prediction — a fan's "vote" is broadcast by the server, not by their own wallet. That's fine for a hackathon MVP (it proves the full MCP → contract → chain path works), but a production version should have the fan sign client-side (e.g. via Keplr) and have the MCP server only build the unsigned tx, or verify a fan-signed payload before relaying it. Worth a sentence in your submission write-up — judges tend to reward projects that name their own known limitations over ones that gloss over them.
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。