Advanced Workflow Mechanics
Master complex development workflows with Claude Code's advanced features for reasoning, context management, and automation.
UltraThink Reasoning Mode
Enhanced reasoning capabilities for complex problem solving.
Activation
# Enable UltraThink for current session
claude-code --ultra-think
# Enable for specific queries
claude-code --think-mode=ultra "Design a microservices architecture"
Reasoning Patterns
Multi-Perspective Analysis
You: How should we handle user authentication?
UltraThink: Analyzing authentication approaches...
Security Perspective:
- JWT tokens provide stateless auth
- OAuth2 for third-party integration
- MFA for sensitive operations
Performance Perspective:
- Token validation overhead
- Session storage considerations
- Caching strategies
Developer Experience Perspective:
- Simple integration APIs
- Clear error messages
- Debugging capabilities
Recommendation: JWT + OAuth2 hybrid approach
Step-by-Step Problem Solving
You: Our API is slow under load
UltraThink: Breaking down performance analysis...
Step 1: Identify bottlenecks
- Database query patterns
- Network latency
- CPU utilization
Step 2: Measure impact
- Response time metrics
- Throughput analysis
- Error rate tracking
Step 3: Optimization strategies
- Query optimization
- Caching layers
- Load balancing
Step 4: Implementation plan
- Priority-based rollout
- A/B testing approach
- Rollback procedures
Reasoning Modes
Deep Analysis Mode
claude-code --think-mode=deep "Analyze this codebase architecture"
Provides:
- Comprehensive system analysis
- Architecture pattern identification
- Technical debt assessment
- Improvement recommendations
Fast Decision Mode
claude-code --think-mode=fast "Choose between React and Vue"
Provides:
- Quick comparative analysis
- Pros/cons summary
- Contextual recommendations
Context Window Optimization
Intelligent Context Management
Claude Code automatically optimizes context usage:
Session Context (8k tokens):
├── Current conversation (2k tokens)
├── Relevant file contents (3k tokens)
├── Project structure (1k tokens)
├── Recent changes (1k tokens)
└── Framework knowledge (1k tokens)
Context Strategies
Hierarchical Context
Level 1: Immediate context (current files, recent messages)
Level 2: Project context (structure, dependencies, configs)
Level 3: Domain context (frameworks, patterns, best practices)
Level 4: Historical context (previous sessions, learnings)
Context Compression
# Enable context compression
claude-code --compress-context
# Manual context reset
claude-code --reset-context
# Context usage stats
claude-code --context-stats
Memory Management Strategies
Selective Memory
{
"memory": {
"retain": [
"project-structure",
"coding-conventions",
"team-preferences"
],
"compress": [
"detailed-implementations",
"verbose-outputs"
],
"discard": [
"temporary-calculations",
"debug-information"
]
}
}
Session Continuity
# Save session state
claude-code --save-session my-project-work
# Resume session
claude-code --resume-session my-project-work
# List saved sessions
claude-code --list-sessions
Token Management Techniques
Efficient Tool Usage
Batch Operations
// Good: Batch file operations
const files = ['src/components/A.tsx', 'src/components/B.tsx'];
readMultipleFiles(files);
// Avoid: Individual operations
files.forEach(file => readFile(file));
Strategic Tool Selection
# For specific files: Use Read tool
claude-code read src/components/Button.tsx
# For pattern matching: Use Grep tool
claude-code grep "useState" --type=tsx
# For open-ended search: Use Task tool
claude-code task "Find all authentication-related code"
Context-Aware Responses
Abbreviated Responses
{
"response_style": {
"verbosity": "minimal",
"code_comments": false,
"explanations": "on_request",
"summaries": "brief"
}
}
Progressive Disclosure
Initial Response: Brief solution outline
User Follow-up: Detailed implementation
User Request: Complete code examples
Workflow Automation Patterns
CI/CD Integration
# .github/workflows/claude-integration.yml
name: Claude Code Integration
on: [push, pull_request]
jobs:
claude-review:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v3
- name: Run Claude Code Review
run: |
claude-code review \
--agent=code-reviewer \
--format=github-comments \
--files=$(git diff --name-only)
Development Environment Setup
#!/bin/bash
# .claude/setup.sh
echo "🔧 Setting up development environment..."
# Install dependencies
npm install
# Setup database
docker-compose up -d postgres
npm run db:migrate
# Configure environment
cp .env.example .env
echo "DATABASE_URL=postgresql://localhost/myapp" >> .env
# Setup git hooks
npx husky install
npx husky add .husky/pre-commit "claude-code hooks run pre-commit"
echo "✅ Environment ready"
Automated Code Review
{
"workflows": {
"code-review": {
"trigger": "git-commit",
"agents": ["code-reviewer", "security-auditor"],
"checks": [
"style-consistency",
"security-vulnerabilities",
"performance-issues",
"test-coverage"
],
"reporting": {
"format": "markdown",
"destination": "review-comments.md"
}
}
}
}
Performance Optimization Workflows
Automated Performance Testing
#!/bin/bash
# scripts/performance-test.sh
echo "🏃 Running performance tests..."
# Start application
npm run start:test &
APP_PID=$!
# Wait for startup
sleep 10
# Run load tests
npx autocannon http://localhost:3000 \
--connections 100 \
--duration 60 \
--latency
# Performance regression check
npm run test:performance
# Cleanup
kill $APP_PID
Bundle Analysis
{
"hooks": {
"build-complete": [
"npm run analyze-bundle",
"check-bundle-size.sh",
"generate-performance-report.sh"
]
}
}
Complex Workflow Examples
Full Feature Development
# Multi-step feature development workflow
claude-code workflow run feature-development \
--feature="user-notifications" \
--agents="architect,test-engineer,reviewer" \
--stages="design,implement,test,review,deploy"
Automated steps:
- Design: Architecture agent creates system design
- Implement: Code generation with best practices
- Test: Test engineer creates comprehensive test suite
- Review: Code reviewer validates implementation
- Deploy: DevOps automator handles deployment
Microservice Development
#!/bin/bash
# workflows/microservice.sh
SERVICE_NAME="$1"
echo "🏗️ Creating microservice: $SERVICE_NAME"
# Generate service structure
claude-setup generate microservice \
--name="$SERVICE_NAME" \
--database=postgresql \
--api=graphql
# Setup testing
claude-testing init \
--framework=jest \
--coverage=90 \
--integration=true
# Configure deployment
claude-devops setup \
--container=docker \
--orchestration=kubernetes \
--monitoring=prometheus
# Initialize repository
git init
git add .
git commit -m "feat: initialize $SERVICE_NAME microservice
🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>"
echo "✅ Microservice $SERVICE_NAME ready for development"
Workflow Debugging
Debug Commands
# Debug workflow execution
claude-code workflow debug feature-development
# Workflow performance profiling
claude-code workflow profile --session-id=abc123
# Step-by-step execution
claude-code workflow run --step-mode feature-development
Workflow Logs
# View workflow logs
claude-code workflow logs
# Filter by workflow type
claude-code workflow logs --type=deployment
# Real-time monitoring
claude-code workflow logs --follow --level=debug
Custom Workflow Creation
Workflow Definition
// workflows/custom-deployment.ts
export const customDeployment = {
name: 'custom-deployment',
description: 'Custom deployment workflow for our stack',
steps: [
{
name: 'pre-flight-checks',
agent: 'devops-automator',
task: 'Validate deployment readiness',
required: true
},
{
name: 'build-and-test',
parallel: [
{ task: 'npm run build', timeout: 300000 },
{ task: 'npm run test:ci', timeout: 600000 }
]
},
{
name: 'deploy',
agent: 'deployment-specialist',
task: 'Deploy to production environment',
approval_required: true
},
{
name: 'post-deployment',
tasks: [
'run smoke tests',
'update monitoring dashboards',
'notify stakeholders'
]
}
]
};
Workflow Registration
# Register custom workflow
claude-code workflow register ./workflows/custom-deployment.ts
# List available workflows
claude-code workflow list
# Run custom workflow
claude-code workflow run custom-deployment
Best Practices
1. Workflow Design
- Clear Stages: Define distinct workflow phases
- Error Handling: Plan for failure scenarios
- Rollback Procedures: Enable quick recovery
- Monitoring: Track workflow execution
2. Context Management
- Minimize Context: Only include relevant information
- Session Planning: Structure conversations efficiently
- Memory Optimization: Use appropriate retention policies
3. Performance
- Parallel Execution: Run independent tasks concurrently
- Caching: Cache expensive operations
- Early Termination: Stop on critical failures
- Resource Limits: Set appropriate timeouts
4. Team Collaboration
- Shared Workflows: Standardize team processes
- Documentation: Document custom workflows
- Version Control: Track workflow changes
- Training: Ensure team understanding
Troubleshooting
Common Issues
- Context Overflow: Reduce conversation length or reset context
- Slow Workflows: Optimize expensive operations
- Hook Failures: Check script permissions and dependencies
- Memory Issues: Adjust retention policies
Performance Monitoring
# Workflow performance metrics
claude-code metrics workflows
# Context usage analysis
claude-code metrics context
# Hook execution times
claude-code metrics hooks
Next Steps
- Explore Performance Optimization
- Learn about Hooks Framework
- See IDE Integrations