super-mcp-server

# Super MCP Server A Lambda-deployable MCP server with embedded SuperDB for data querying. ## Features - **Embedded SuperDB**: No external dependencies, Super binary is embedded in the Go executable - **Lambda Ready**: Runs in AWS Lambda with automatic binary extraction to `/tmp` - **Local Development**: Also works as a local MCP server for testing with Claude Desktop - **Three Core Tools**: - `query_data`: Execute SuperSQL queries on data files or URLs - `list_files`: List data files in directories - `analyze_data`: Analyze data structure and types ## Setup 1. **Download Super binary**: ```bash chmod +x scripts/setup.sh ./scripts/setup.sh ``` 2. **Build for local testing**: ```bash go mod tidy go build -o super-mcp-server ``` 3. **Build for Lambda**: ```bash GOOS=linux GOARCH=amd64 go build -o bootstrap main.go zip lambda-deployment.zip bootstrap ``` ## Local Usage Configure Claude Desktop (`claude_desktop_config.json`): ```json { "mcpServers": { "super-data": { "command": "/absolute/path/to/super-mcp-server", "args": [], "env": {} } } } ``` Then restart Claude Desktop and you can ask questions like: - "List the data files in my current directory" - "Query my sample.json file and show the first 10 records" - "Analyze the structure of this CSV file" ## Lambda Deployment 1. Upload `lambda-deployment.zip` to AWS Lambda 2. Set runtime to "Provide your own bootstrap on Amazon Linux 2" 3. Set handler to "bootstrap" 4. Increase timeout (recommend 30+ seconds for large queries) 5. Increase memory (recommend 512MB+ for better performance) ## Project Structure ``` super-mcp-server/ ├── main.go # Complete server implementation ├── go.mod # Go module dependencies ├── scripts/ │ └── setup.sh # Downloads Super binary ├── binaries/ │ └── super-linux-amd64 # Embedded Super binary (created by setup.sh) └── README.md # This file ``` ## Example Queries Once connected to Claude Desktop, you can ask: ### Basic Queries - "What data files do I have in my current directory?" - "Show me the first 5 records from data.json" - "Count the rows in my CSV file" ### Advanced Analytics - "Find the top 10 users by activity in my logs" - "Group sales data by month and show totals" - "Analyze the data types in my JSON file" ### URL Data Sources - "Query this API endpoint and show me the structure: https://api.example.com/data.json" - "Download and analyze data from this URL" ## SuperSQL Language The server uses SuperDB's SuperSQL dialect, which extends SQL with: - **Pipe syntax**: `SELECT * FROM data.json | WHERE age > 25 | GROUP BY department` - **JSON native**: No need for special JSON functions - **Schema-less**: Works with heterogeneous data - **Rich types**: Supports complex nested data structures ## Technical Notes - The Super binary is embedded at compile time (~15-20MB) - In Lambda, the binary is extracted to `/tmp` on first invocation - Supports data from files, URLs, S3 paths (if Lambda has permissions) - Works with JSON, CSV, Parquet, NDJSON, and many other formats - Container reuse means binary extraction only happens once per container lifecycle ## Troubleshooting ### "Super binary not found" error - Make sure you ran `./scripts/setup.sh` before building - Check that `binaries/super-linux-amd64` exists and is executable ### Claude Desktop connection issues - Use absolute paths in the configuration - Restart Claude Desktop after configuration changes - Check that the binary is executable: `chmod +x super-mcp-server` ### Lambda timeout errors - Increase Lambda timeout for large data processing - Consider using Lambda layers for very large deployments - Monitor CloudWatch logs for specific error messages