MCP Time Server Node

# 9. Rate Limiting in MCP Servers (Verified 2025-07-19) ## Key Findings: 1. **Client Identification**: - MCP servers with stdio transport have 1:1 client relationship - No client identifiers in JSON-RPC requests - Rate limiting is per-server-instance 2. **Error Response Format**: - Use JSON-RPC error code -32000 (server-defined error) - Include data with limit, window, and retryAfter ```typescript { error: { code: -32000, message: 'Rate limit exceeded', data: { limit: 100, window: 60000, retryAfter: 45 } } } ``` 3. **Integration Point**: - Add rate limit check in CallToolRequestSchema handler - Check before tool execution to avoid unnecessary processing - Don't rate limit tools/list requests 4. **Algorithm**: - Sliding window counter provides fairness - Tracks request timestamps in memory - Automatic cleanup of expired timestamps 5. **Configuration**: - Environment variables: RATE_LIMIT, RATE_LIMIT_WINDOW - Defaults: 100 requests per 60000ms (1 minute)