carrot-ai-pm

Carrot AI PM acts as an AI coding project manager that helps you specify, validate, and align AI-generated code with detailed specifications. It integrates with your AI coding assistant workflow to create clear specs for APIs, UI components, databases, and CLI tools, then checks that the produced code adheres to those specs. The system can validate endpoints, security requirements, input validation, and error handling, and it provides actionable suggestions to bring implementations in line with the defined specification. In practice, you would define a spec for a given artifact (e.g., a REST API or a UI component), have your AI assistant implement the code based on that spec, and then run Carrot AI PM to verify compliance and surface any gaps or fixes needed before moving forward.

Prerequisites:

• Node.js 18+ installed
• npm (comes with Node.js) or pnpm/yarn if preferred
• Git

1. Clone the repository or install via npm (depending on your setup):

# Optional: clone the repository
# git clone https://github.com/talvinder/carrot-ai-pm.git
# cd carrot-ai-pm

# Install dependencies
npm install

2. Build the project (if there is a build step):

npm run build

3. Run the MCP server locally (example):

# Ensure the server entry point exists, typically dist/src/server.js
node dist/src/server.js

4. Configure your AI assistant integration (example for Cursor):

• Create or edit .cursor/mcp.json as shown in the Quick Setup section of the README, pointing to the built server entry point and any required environment variables.

{
  "mcpServers": {
    "carrot-pm": {
      "command": "node",
      "args": ["/path/to/carrot-ai-pm/dist/src/server.js"],
      "env": {
        "CARROT_PROJECT_ROOT": "/path/to/your/project"
      }
    }
  }
}

Tips and common considerations: - Ensure you’re using Node.js 18+ for compatibility with modern tooling. - Keep your CARROT_PROJECT_ROOT and other paths updated to reflect your local environment. - If you see mismatches in validation, review the specific checks Carrot performs (structure, security, validation, documentation) and iterate on the spec before re-running validation. - Carrot operates locally and emphasizes static analysis; there is no code execution. - For best results, run Carrot after each meaningful change to your spec or implementation to catch regressions early.