Skip to main content

AI-Assisted Development

NestForge Pro is designed from the ground up to work well with AI coding agents such as Claude Code, Cursor, GitHub Copilot, and similar tools.

CLAUDE.md​

The CLAUDE.md file at the project root provides AI agents with complete context about the codebase: the tech stack, repository structure, module priority, Prisma schema, coding standards, and environment variable reference.

When starting a session with an AI coding agent, point it at this file. Most tools pick it up automatically when it exists in the project root.

Why NestForge works well with AI​

Predictable architecture​

Every domain module follows the same pattern:

modules/
└── <domain>/
    ├── <domain>.module.ts       # NestJS module definition
    ├── <domain>.controller.ts   # HTTP layer (routes, validation, Swagger)
    ├── <domain>.service.ts      # Business logic
    ├── <domain>.service.spec.ts # Unit tests
    └── dto/                     # Request/response DTOs with class-validator decorators

AI agents perform best when patterns are consistent. Once the agent understands one module, it can work with any module in the project.

Explicit interfaces and named exports​

API contracts are defined as TypeScript interfaces and constants in packages/shared/src/. DTOs use class-validator decorators for runtime validation; Zod is used only for environment config. There are no implicit types or magic string conventions for the agent to guess at.

Co-located types​

Types live next to the code that uses them. Request/response DTOs sit in a dto/ folder inside each module. Shared types and plan constants live in packages/shared/src/. The agent does not need to search across unrelated directories to find type definitions.

Swagger documentation​

Every API endpoint has Swagger decorators. The generated documentation at /api/docs serves as a machine-readable API reference that agents can use to understand request/response shapes without reading implementation code.

Tips for using AI with NestForge​

  1. Point your agent at CLAUDE.md -- This gives it the full project context in one read. Claude Code and Cursor load it automatically.

  2. Use the Swagger docs -- The interactive API documentation at http://localhost:4000/api/docs is useful both for humans and as a reference for AI agents when working on frontend integration.

  3. Follow the module pattern -- When asking an AI agent to create a new feature, reference an existing module (e.g., "follow the same pattern as modules/teams"). The consistent structure makes generation more reliable.

  4. Leverage the shared package -- When defining new API contracts, add types to packages/shared/src/ so both the API and frontend stay in sync. AI agents handle this well when they can see the existing type definitions.

  5. Run tests after changes -- Ask the agent to run npx turbo test after making changes. The test suite catches regressions quickly and gives the agent feedback to self-correct.