Skip to content

AI Development Workflows

AI Velocity Exclusive

AI development features are exclusive to AI Velocity tier.

Best practices for AI-assisted development.

Using Cursor

Basic Workflow

  1. Open file you want to modify
  2. Start typing - Cursor suggests based on rules
  3. Accept suggestions with Tab
  4. Reference rules with @rule-name if needed

Generating New Code

Text Only
1
2
3
4
5
6
7
8
# In chat:
"Create a new API endpoint for user statistics following the clean architecture pattern"

# Cursor knows:
- Clean architecture (Controllers → Services → Models)
- FastAPI patterns
- Where to put files
- How to structure code

Refactoring

Text Only
1
2
3
"Refactor this service to use dependency injection"

# Cursor applies patterns from rules automatically

Using Claude Code

Basic Workflow

Bash
1
2
cd your-project
claude
Text Only
1
2
3
4
5
6
7
8
9
# Ask questions
"How does the payment system work?"

# Generate code
/new-api-endpoint GET /api/users/{id}/activity

# Run tasks
/run-tests
/fix-lint

Multi-Claude Workflow

For complex features, run multiple Claude instances:

Bash
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
# Terminal 1: Backend
claude
# "Implement notification system backend"

# Terminal 2: Frontend
claude  
# "Create notification UI components"

# Terminal 3: Testing
claude
# "Write tests for notifications"

Each Claude works independently!

Best Practices

DO:

  • ✅ Let AI read context files first
  • ✅ Be specific with instructions
  • ✅ Use custom commands for common tasks
  • ✅ Review generated code before accepting
  • ✅ Run tests after AI changes
  • ✅ Use /clear between unrelated tasks (Claude)

DON'T:

  • ❌ Ask AI to do everything at once
  • ❌ Skip code review
  • ❌ Forget to regenerate API client
  • ❌ Mix multiple unrelated features
  • ❌ Disable safety checks

Common Workflows

Adding New Feature

With Cursor: 1. Open relevant files 2. Describe feature in chat 3. Let Cursor generate following rules 4. Review and refine 5. Run tests

With Claude:

Text Only
1
2
3
/new-api-endpoint POST /api/feature creates new feature
/new-component FeatureCard displays feature info
/run-tests

Debugging Issues

With Cursor: 1. Paste error message 2. Ask "What's causing this error?" 3. Let Cursor analyze and suggest fix 4. Apply fix 5. Test

With Claude:

Text Only
1
2
3
"The payment webhook is failing with 400 error. 
Here's the error log: [paste log]
Analyze and fix."

Refactoring Code

With Cursor:

Text Only
1
"Refactor this service to use the repository pattern from the backend architecture rules"

With Claude:

Text Only
1
2
"Refactor the analytics service to improve performance.
Follow the patterns in CLAUDE.md."

AI Context Tips

What AI Knows (from CLAUDE.md & Rules)

  • ✅ Project architecture
  • ✅ Tech stack
  • ✅ Common commands
  • ✅ Code patterns
  • ✅ File structure
  • ✅ Best practices
  • ✅ Common pitfalls

What to Tell AI

  • What feature you're building
  • Specific requirements
  • Constraints or preferences
  • Files to modify

What AI Doesn't Know

  • Your specific business logic
  • External API details (provide docs)
  • Custom requirements
  • Non-standard patterns

Productivity Tips

Use Templates (Cursor Rules)

AI knows templates for: - API endpoints - React components - Database migrations - Service classes - Controller endpoints

Just describe what you want!

Use Commands (Claude Code)

For repetitive tasks: - /new-api-endpoint - Generate endpoint - /new-component - Create component - /create-migration - Database migration - /run-tests - Execute tests - /fix-lint - Auto-fix linting

Iterative Refinement

Text Only
1
2
3
4
5
6
7
8
You: "Create a user dashboard"
AI: [generates basic dashboard]

You: "Add charts showing user activity"
AI: [adds charts]

You: "Make it responsive for mobile"
AI: [adds responsive design]