Skip to content

Project Setup

Setting up a development project effectively using AI tools and techniques ensures smooth collaboration, high productivity, and adherence to best practices. Below is a detailed guide for initializing your project:

This guide outlines how to set up a development project using AI tools and techniques. A proper setup ensures: - smooth collaboration - high productivity - adherence to best practices

Initial Setup

  1. Choose Project Platform

    • select hosting platform (Azure DevOps, GitHub, GitLab, Bitbucket)
    • agree on Git branching strategy
    • document collaboration guidelines

  2. Configure Repository

    • create GitHub repository manually
    • set branch protection rules

Development Environment

1. Choose Your AI IDE or Extension

Select an AI-powered development environment that fits your workflow:

  • Cursor - AI-first code editor with built-in AI chat and editing (subscription required)
  • Kiro - AI IDE with steering documents for consistent AI behavior (subscription required)
  • GitHub Copilot - AI pair programmer available as VS Code/JetBrains plugin (subscription required)
  • Claude Code - AI coding agent by Anthropic, available as CLI and IDE extension (Claude Pro subscription required)
  • OpenAI Codex - AI coding agent by OpenAI, available as cloud service, CLI tool, and IDE extensions (subscription required)

2. Install IDE and Required Plugins

For Cursor and Kiro:

  • Download and install the IDE (both are built on VS Code)
  • Install required VS Code extensions for your tech stack
  • Configure workspace settings and preferences

For GitHub Copilot:

  • Install GitHub Copilot extension in VS Code or JetBrains IDE
  • Sign in with GitHub account

For Claude Code:

For OpenAI Codex:

  • Install IDE extension from marketplace or use CLI tool
  • Configure API access with OpenAI API key
  • See Codex IDE Configuration for setup details

3. Configure Privacy Settings

IMPORTANT: Privacy settings must be enabled. This ensures that data is not persisted on providers' servers and is not used for training models.

  • ChatGPT: Settings -> Data Controls -> "Improve the model for everyone" -> toggle off/disable
  • Cursor: Settings -> General -> Privacy Mode -> enabled
  • Kiro: Configure privacy settings according to your organization's policy
  • GitHub Copilot: Settings -> Data retention -> Configure according to your organization's policy
  • Claude Code: Configure via settings files
  • OpenAI Codex: TBA

4. Configure IDE Rules

IDE rules guide AI behavior and provide context about your project. Learn more in IDE Rules Overview.

Common Steps for All IDEs:

  • Copy language-specific rules from programming language rules
  • Copy framework-specific rules from framework rules
  • Customize rules for your team's requirements and coding standards
  • Update system prompt for LLM base understanding

IDE-Specific Rule Configuration:

For Cursor:

  • Enable Rules & Memories in Cursor Settings:
    • Go to Settings -> Features -> Rules & Memories
    • Enable "Include CLAUDE.md in context" - adds CLAUDE.md and CLAUDE.local.md files to the agent's context when relevant
    • Enable "Import Claude Commands" - loads commands from .claude/commands directories alongside .cursor/commands
  • Store rules in .cursor/rules/ directory using MDC format
    • Support for Always, Auto Attached, Agent Requested, and Manual rule types

For Kiro:

  • Store steering documents in .kiro/steering/ directory
    • Configure file match patterns for conditional inclusion
    • Support for Always Included, File Match Conditional, and Manual Inclusion

For Claude Code:

  • User-level:
    • Create ~/CLAUDE.md file for custom user prompts
    • Create ~/.claude/settings.json for global configurations
    • Store user commands in ~/.claude/commands/ directory (user-level)
  • Project-level:
    • Create CLAUDE.md for the current repository/folder using /init command
    • Create .claude/settings.json for project-level configuration (shared with team)
    • Create .claude/settings.local.json for local settings (not checked into source control)
    • Store commands in .claude/commands/ directory (project-level)
  • Configure permissions, environment variables, and hooks in settings files
  • See Claude Code Settings documentation for complete configuration options

For OpenAI Codex:

  • Configure IDE-specific settings according to Codex IDE configuration
  • Set up API key and model preferences
  • Create project-specific configuration files as needed

5. Set Up Slash Commands

Slash commands reduce token consumption and enable reusable workflows. Learn more in Slash Commands Guide.

For Cursor:

  • Commands stored in .cursor/commands/ directory
  • Also supports .claude/commands/ when "Import Claude Commands" is enabled

For Claude Code:

  • Project commands in .claude/commands/ - appear as (project) in /help
  • Personal commands in ~/.claude/commands/ - appear as (user) in /help
  • Create Markdown files - filename becomes the command name
  • Support for namespacing with subdirectories

Using Repository Documentation

Create directory within the repository to store documentation, such as:

  • product requirements
  • architecture documentation
    • development rules and guidelines (reference IDE rules)

Ensure all documentation is:

  • version controlled
  • accessible to the IDE
  • formatted in markdown

(Optinal) Initialize Memory Bank with Tools

Set up the Memory Bank system to enable effective AI assistance throughout the project lifecycle:

  1. Create Memory Bank Structure
    • Install Memory Bank MCP
    • Initialize core memory bank files using prompt: Initialize memory bank with tools
  2. Configure Memory Bank Files
    • projectbrief.md - foundation document defining project scope and goals
    • productContext.md - user experience goals and problem definition
    • activeContext.md - current work focus and recent changes
    • systemPatterns.md - architecture and technical decisions
    • techContext.md - technologies and development setup
    • progress.md - current status and evolution tracking
  3. Integrate with Development Workflow
    • Ensure memory bank files are version controlled
    • Establish update patterns for maintaining current context
    • Configure AI tools to reference memory bank for project understanding

The Memory Bank serves as the AI assistant's primary reference for understanding project context, enabling more effective and contextually-aware assistance throughout development.

(Optional) Configure MCP Server for Spec-Driven Development

Set up the VibeSpecs MCP Server to enable structured, spec-driven development workflow in Cursor / VS Code:

  1. Install VibeSpecs MCP Server
    • Follow installation instructions from Spec-Diven Dev MCP
    • Configure the server in your Cursor MCP settings
  2. Initialize Spec-Driven Workflow
    • Use the workflow commands to guide feature development from requirements to implementation
    • Leverage structured phases: goal definition → requirements → design → tasks → execution
  3. Integrate with Project Structure
    • Align spec outputs with your repository documentation structure
    • Ensure generated specifications are version controlled
    • Reference Spec-Driven Development for best practices

The Spec-Driven Dev MCP Server provides a systematic approach to feature development, ensuring thorough planning and consistent implementation across your project.

Next Steps

Once your project setup is complete, proceed to the feature development lifecycle:

Requirement Specification - Start by defining what your system should do from a user's perspective, including expected behaviors, UI elements, and testable features.