How richardsondx/checkmate works
Project type and purpose
• CheckMate is a Node.js/TypeScript CLI application and AI-powered Test-Driven-Development (TDD) framework.
• Its purpose is to generate, run, and verify “specs” (Markdown or YAML) against your code using LLMs (OpenAI & Anthropic), with optional in-editor integration via Cursor’s Middleware Control Protocol (MCP).
File structure
• bin/ – Platform-agnostic CLI “wrapper” scripts (checkmate, checkmateai)
• src/
• commands/ – one file per CLI command (init, gen, status, watch, etc.)
• lib/ – domain libraries (AI client, file I/O, caching, telemetry, spec parsing/validation)
• mcp/ – MCP server entrypoint & router for Cursor integration
• ui/ – ASCII banner & markdown rendering utilities
• index.ts – CLI dispatcher
• scripts/ – project-scaffolding and maintenance scripts (cursor rules, spec templates)
• .cursor/ – Cursor AI configuration & injected rule files (.mdc)
• dist-test/ – compiled artifacts used by “bin” (mirrors dist/) for end-to-end test harness
• memory-docs/ & wiki/ – design docs, guides, standards
• tests/ – unit & integration tests
• package.json, tsconfig.json, .checkmate-config.json – build & runtime configuration
README for additional insights
• Dual-model architecture: uses “reason” (Claude) for spec generation and “quick” (GPT) for verification
• Smart caching in .checkmate/cache/ to minimize API calls
• .checkmate YAML or env variables for config
• MCP server registers under .cursor/config.json to enable live in-editor commands
• Visual indicators in Cursor driven by injected rule files
Diagramming conventions
• Blue for “CLI & Commands”
• Green for “Domain Services”
• Orange for “Infrastructure (FS, Cache, Logs)”
• Purple for “External Services (OpenAI, Anthropic)”
• Gray for “Editor Integration (MCP ↔ Cursor)”
• Rectangles for modules/components
• Double-lined for in-process IPC (MCP channel)