DeepWriter MCP Server

# System Patterns: DeepWriter MCP Server ## Architecture Overview - **MCP Server**: Implements stdio or SSE protocol for tool communication. - **Tool Layer**: Each DeepWriter API operation is exposed as a discrete MCP tool with strict input/output schemas. - **API Integration**: Uses HTTP(S) requests to DeepWriter endpoints, passing the API key in the `x-api-key` header. - **Error Handling**: Centralized error handler for mapping API errors (401, 403, 404, 400, 500) to MCP tool responses. - **Security**: API keys are never logged; all requests use HTTPS; input is sanitized; rate limiting is enforced. - **Logging**: Activity and errors are logged (excluding sensitive data) for debugging and transparency. - **Resource Layer**: Exposes API key validation status, rate limits, usage stats, error logs, and request history as MCP resources. ## Key Technical Decisions - Use Node.js/TypeScript for rapid development and strong typing. - Use `fetch` for HTTP requests. - Use `dotenv` or environment variables for secure API key storage. - Implement retry logic for transient API failures. - Modularize each tool for maintainability and testability. ## Design Patterns - **Command Pattern**: Each MCP tool is a command handler. - **Adapter Pattern**: Adapts DeepWriter API responses to MCP tool output schemas. - **Singleton/Context**: Centralized context for logging, rate limiting, and resource management. ## Component Relationships - MCP server core manages tool registration and protocol. - Each tool module interacts with the DeepWriter API and error handler. - Logging and resource modules are shared across tools. ## Extensibility - New DeepWriter endpoints can be added as new MCP tools. - Schemas and error handling are easily updatable.