FTC Platform MCP Server
A standalone MCP server that provides AI tools by proxying requests to the FTC Platform's REST API, enabling tasks like fetching event applications, evaluations, and scoring criteria.
README
FTC Platform MCP Server
A standalone Model Context Protocol (MCP) server for the FTC Platform that provides AI tools by proxying requests to the main platform's REST API.
Architecture
This MCP server runs independently from the main Next.js application and communicates with it via HTTP API calls:
┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐
│ Claude AI │───▶│ MCP Server │───▶│ Vercel API │
│ (MCP Client) │ │ (This App) │ │ (ftc-platform) │
└─────────────────┘ └─────────────────┘ └─────────────────┘
Why Separate Deployment?
- Vercel Limitations: Vercel Functions don't support WebSocket servers required for MCP
- Long-lived Connections: MCP sessions are persistent and bidirectional
- Independent Scaling: MCP server can scale separately from the main app
- Clean Architecture: Separation of concerns between web app and AI tools
Available Tools
- test_connection - Verify MCP server and API connectivity
- get_event_applications - Fetch all applications for an event with complete data
- get_event_evaluations - Get completed evaluations with scores and statistics
- get_evaluation_criteria - Get scoring criteria categorized for AI understanding
- get_application_questions - Get application question structure and metadata
Environment Variables
Create a .env file from .env.example:
cp .env.example .env
Required variables:
VERCEL_API_BASE_URL- Base URL of your deployed FTC Platform APIMASTRA_API_KEY- API key for authenticating with the Vercel API
Development
Prerequisites
- Node.js 20+
- Access to the FTC Platform API endpoints
Local Development
- Install dependencies:
npm install
- Set up environment variables:
cp .env.example .env
# Edit .env with your actual values
- Start development server:
npm run dev
- Test the connection:
# In another terminal, test via stdio
echo '{"jsonrpc": "2.0", "id": 1, "method": "tools/list"}' | npm run dev
Deployment
Fly.io (Recommended)
- Install Fly CLI:
curl -L https://fly.io/install.sh | sh
- Login and create app:
fly auth login
fly apps create ftc-platform-mcp
- Set environment variables:
fly secrets set VERCEL_API_BASE_URL="https://your-app.vercel.app/api/mastra"
fly secrets set MASTRA_API_KEY="your-secret-key"
- Deploy:
fly deploy
Docker (Alternative)
# Build
docker build -t ftc-platform-mcp .
# Run locally
docker run -p 3001:3001 \
-e VERCEL_API_BASE_URL="https://your-app.vercel.app/api/mastra" \
-e MASTRA_API_KEY="your-secret-key" \
ftc-platform-mcp
Claude Configuration
After deploying, configure Claude to use your MCP server:
// .mcp.json
{
"mcpServers": {
"ftc-platform": {
"type": "stdio",
"command": "ssh",
"args": [
"your-server.fly.dev",
"node",
"/app/dist/server.js"
],
"env": {
"NODE_ENV": "production"
}
}
}
}
API Authentication
The MCP server authenticates with the Vercel API using the same MASTRA_API_KEY that the main platform uses. This provides:
- Reuse of existing auth - No changes needed to Vercel API
- Server-to-server security - API key is stored securely on the MCP server
- Access control - Same permissions as the main platform
Project Structure
src/
├── server.ts # Main MCP server implementation
├── lib/
│ └── api-client.ts # HTTP client for Vercel API calls
└── types/
└── index.ts # TypeScript type definitions
# Deployment configs
├── Dockerfile # Docker container definition
├── fly.toml # Fly.io deployment config
└── README.md # This file
Monitoring
Health Checks
- HTTP endpoint:
http://your-app:3001/health - Fly.io automatically monitors and restarts unhealthy instances
Logs
# Fly.io logs
fly logs
# Docker logs
docker logs container-id
Troubleshooting
Common Issues
-
Connection Failed
- Check
VERCEL_API_BASE_URLis correct and accessible - Verify
MASTRA_API_KEYis valid
- Check
-
Tool Execution Errors
- Check Vercel API is responding correctly
- Verify eventId parameters are valid UUIDs
-
MCP Client Issues
- Ensure Claude has the correct MCP server configuration
- Check stdio communication is working properly
Debug Mode
Set environment variable for verbose logging:
NODE_ENV=development npm run dev
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。