• Generates structured, enforceable specs (OpenAPI, UI, DB, CLI)
• Runs AST-level checks to verify the AI implementation matches the spec
• Flags skipped validation, missing auth, hallucinated functions, etc.
• Stores every result (.carrot/compliance/*) for audit trails and trend tracking
• 100 % local – Carrot never calls external APIs; it just uses Cursor’s internal LLM hook
Quick start: git clone https://github.com/talvinder/carrot-ai-pm npm install && npm run build # add to .cursor/mcp.json, then chat: # “Create spec for X → implement → check implementation”
Example output: endpoints match missing rate-limit on POST /login undefined fn: sendWelcomeEmail()
I built it after years of watching markdown specs drift. Edge-cases, bugs, and PRs welcome → https://github.com/talvinder/carrot-ai-pm/issues
talvinder•8mo ago
• Generates structured, enforceable specs (OpenAPI, UI, DB, CLI)
• Runs AST-level checks to verify the AI implementation matches the spec
• Flags skipped validation, missing auth, hallucinated functions, etc.
• Stores every result (.carrot/compliance/*) for audit trails and trend tracking
• 100 % local – Carrot never calls external APIs; it just uses Cursor’s internal LLM hook
Quick start: git clone https://github.com/talvinder/carrot-ai-pm npm install && npm run build # add to .cursor/mcp.json, then chat: # “Create spec for X → implement → check implementation”
Example output: endpoints match missing rate-limit on POST /login undefined fn: sendWelcomeEmail()
I built it after years of watching markdown specs drift. Edge-cases, bugs, and PRs welcome → https://github.com/talvinder/carrot-ai-pm/issues