An MCP server designed exclusively for AI assistants that enables querying Tideways performance monitoring data for PHP applications. It exposes a set of MCP tools to retrieve aggregate performance metrics, time-series summaries, recent issues, traces, and historical data directly as raw JSON, suitable for LLM analysis without any formatting. The server talks to Tideways via a dedicated HTTP client with built-in rate limiting, automatic retries, and secure token handling, configurable by environment variables (token, organization, project, base URL, retries, timeout, log level). Its architecture includes the main MCP server implementation, a Tideways API client, a modular tool registry with individual handlers, and centralized error handling and logging. Data flow is AI Assistant ← MCP (stdio) ← Tideways MCP Server → Tideways API, with raw responses delivered unchanged to the AI. Available tools include get_performance_metrics, get_performance_summary, get_issues, get_traces, and get_historical_data, enabling conversational prompts like “What are the current performance metrics?” and “Show me slow traces.” The server is designed to integrate with MCP-compatible assistants (Claude Desktop, Cursor, VS Code MCP extension) and uses a simplified raw JSON response format for optimal performance.