Claude Code Headless

1
2
3
4
5

# Install Claude Code
npm install -g @anthropic/claude-code

# Set up your API key
export ANTHROPIC_API_KEY=your_api_key_here

1

claude -p "Your prompt here"

1
2
3
4
5

# Get JSON output with metadata
claude -p "Review this file" --output-format json

# Stream JSON for real-time processing
claude -p "Long running task" --output-format stream-json

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13

# Analyze system logs for anomalies
claude -p "Analyze the last 1000 lines of /var/log/nginx/access.log \
  and identify unusual traffic patterns or potential security threats" \
  --allowedTools "Read,Grep"

# Generate infrastructure health reports
claude -p "Create a comprehensive health report based on CPU, memory, \
  and disk usage metrics in the monitoring dashboard" \
  --output-format json

# Predictive maintenance alerts
claude -p "Based on recent performance metrics, predict potential \
  system bottlenecks and suggest preventive actions"

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

# Daily KPI analysis
claude -p "Analyze yesterday's sales data and generate executive \
  summary with trends and recommendations" --output-format json

# Competitive intelligence reports
claude -p "Review competitor pricing updates from scraped data \
  and suggest pricing strategy adjustments"

# Customer behavior analysis
claude -p "Process user interaction logs and identify friction \
  points in the customer journey"

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12

# SEO content optimization
claude -p "Review and optimize website content in the /content \
  directory for SEO best practices and readability" \
  --allowedTools "Read,Edit"

# Social media automation
claude -p "Generate social media content calendar based on recent \
  blog posts and company announcements" --output-format json

# Image and asset analysis
claude -p "Audit all images in /assets for accessibility, \
  optimization, and compliance with brand guidelines"

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

# Contract analysis
claude -p "Review vendor contracts in /legal/contracts and flag \
  potential risk clauses or compliance issues" \
  --allowedTools "Read,Grep"

# Privacy policy updates
claude -p "Update privacy policy to reflect new data collection \
  practices while maintaining GDPR compliance"

# Audit trail generation
claude -p "Generate compliance audit report based on system logs and user activity data" --output-format json

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

# Deployment risk assessment
claude -p "Analyze the planned deployment changes and assess \
  potential risks, suggesting rollback strategies"

# Infrastructure cost optimization
claude -p "Review cloud resource usage patterns and recommend cost optimization strategies" --output-format json

# Incident response automation
claude -p "Based on current system alerts, prioritize issues \
  and generate incident response playbook"

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17

# Comprehensive security audit with automatic issue creation
claude --model opus -p "I would like you to find high and critical \
  security issues in this codebase. Create the issues on github directly. \
  Please avoid false positives and use descriptive title and descriptions" \
  --allowedTools "Bash"

# Security log analysis
claude -p "Analyze security logs for the past 24 hours and identify \
  potential threats or breach attempts" --allowedTools "Read,Grep"

# Vulnerability assessment
claude -p "Review recent dependency updates and assess security \
  implications with mitigation strategies"

# Access pattern analysis
claude -p "Analyze user access patterns and flag suspicious login \
  activities or privilege escalations"

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

# Data quality assessment
claude -p "Analyze data quality in the user database and identify \
  inconsistencies or missing critical information"

# ETL process optimization
claude -p "Review data transformation scripts and suggest \
  performance improvements and error handling"

# Data lineage documentation
claude -p "Generate documentation for data flow from source systems to analytics warehouse" --output-format json

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

# Support ticket analysis
claude -p "Analyze support tickets from the past week and identify \
  common issues and resolution patterns"

# Knowledge base optimization
claude -p "Review support documentation and suggest improvements \
  based on recent customer queries"

# Escalation prediction
claude -p "Based on ticket content and customer history, predict \
  which issues might require escalation"

1
2
3
4
5

# Start a session
SESSION_ID=$(claude -p "Start code review" --output-format json | jq -r '.sessionId')

# Continue the conversation
claude -p "Focus on security issues" --resume $SESSION_ID

1
2
3
4
5
6

# Only allow read operations
claude -p "Review the codebase" --allowedTools "Read,Grep,Glob"

# Custom system prompt for specialized behavior
claude -p "Review for accessibility" \
  --append-system-prompt "Focus specifically on WCAG compliance"

1
2
3
4

# Parse results programmatically
RESULT=$(claude -p "Check test coverage" --output-format json)
echo $RESULT | jq '.content'  # Get the response
echo $RESULT | jq '.cost'     # Get cost information

1
2
3
4
5
6

# Error handling example
if ! claude -p "Run tests and fix failures" --output-format json > result.json 2>&1; then
    echo "Claude Code execution failed"
    cat result.json
    exit 1
fi

1
2
3

claude -p "Refactor large codebase" --output-format stream-json | while read line; do
    echo "Progress: $line"
done

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25

name: Daily Business Analytics
on:
  schedule:
    - cron: '0 8 * * *'  # Run at 8 AM daily
jobs:
  business-insights:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v2
      - name: Install Claude Code
        run: npm install -g @anthropic/claude-code
      - name: Generate Business Insights
        env:
          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
        run: |
          DATE=$(date +%Y-%m-%d)
          claude -p "Analyze yesterday's business metrics and generate \
            executive summary with actionable insights" \
            --output-format json > insights-$DATE.json
          
          # Send alert if revenue dropped significantly
          ALERT=$(jq -r '.alerts.revenue' insights-$DATE.json)
          if [[ "$ALERT" == "critical" ]]; then
            echo "Revenue alert detected" >> $GITHUB_STEP_SUMMARY
          fi

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15

#!/bin/sh
# .git/hooks/pre-commit
# Check content changes for compliance issues

git diff --cached --name-only | grep -E '\.(md|txt|html)$' | while read file; do
  claude -p "Review $file for brand compliance, legal risks, and content quality" \
    --allowedTools "Read" \
    --output-format json > "/tmp/compliance-$file.json"
  
  RISK=$(jq -r '.risk_level' "/tmp/compliance-$file.json")
  if [[ "$RISK" == "high" ]]; then
    echo "ERROR: High-risk content detected in $file"
    exit 1
  fi
done

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16

# /etc/cron.d/infrastructure-health
# Run infrastructure health checks every hour
0 * * * * root /usr/local/bin/infrastructure-monitor.sh

#!/bin/bash
# infrastructure-monitor.sh
claude -p "Analyze system performance metrics and generate health \
  report with predictive maintenance recommendations" \
  --output-format json --allowedTools "Read,Bash" > \
  "/var/log/health-$(date +%H).json"

# Check for critical issues
CRITICAL=$(jq -r '.critical_issues' "/var/log/health-$(date +%H).json")
if [[ "$CRITICAL" != "null" ]]; then
  echo "$CRITICAL" | mail -s "Infrastructure Alert" ops@company.com
fi

1
2
3
4
5
6
7

# Review multiple components
claude -p "Review all React components in src/components/ for accessibility compliance"

# Batch refactoring with controlled tools
claude -p "Refactor all utility functions to use consistent naming \
  conventions and add proper error handling" \
  --allowedTools "Read,Edit,MultiEdit"

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

# Start a feature implementation session
SESSION_ID=$(claude -p "Plan implementation of user authentication feature" --output-format json | jq -r '.sessionId')

# Continue with database models
claude -p "Now implement the database models" --resume $SESSION_ID

# Add API endpoints
claude -p "Create the API endpoints" --resume $SESSION_ID

# Finalize with tests
claude -p "Generate comprehensive tests for the authentication flow" --resume $SESSION_ID

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15

#!/bin/bash
# daily-insights.sh
DATE=$(date +%Y-%m-%d)

# Generate daily business insights
claude -p "Analyze sales data from $DATE and create executive \
  dashboard with KPIs, trends, and recommendations" \
  --output-format json --allowedTools "Read,Grep" > \
  reports/daily-insights-$DATE.json

# Extract key metrics for alerts
REVENUE=$(jq '.revenue_trend' reports/daily-insights-$DATE.json)
if [[ "$REVENUE" == *"declining"* ]]; then
  echo "ALERT: Revenue declining trend detected" | mail -s "Business Alert" executives@company.com
fi

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14

#!/bin/bash
# cost-optimizer.sh
# Analyze cloud usage and suggest optimizations

claude -p "Review AWS usage patterns from CloudWatch metrics \
  and recommend cost optimization strategies" \
  --output-format stream-json --allowedTools "Read,Bash" | \
  while read line; do
    echo "$(date): $line" >> cost-analysis.log
    # Check for urgent recommendations
    if echo "$line" | grep -q "URGENT"; then
      echo "$line" | mail -s "Urgent Cost Alert" devops@company.com
    fi
  done

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16

#!/bin/bash
# compliance-check.sh
# Automated content review for legal and brand compliance

find ./content -name "*.md" -type f | while read file; do
  claude -p "Review content in $file for GDPR compliance, brand \
    guidelines, and legal risks" --allowedTools "Read" \
    --output-format json > \
    "compliance-reports/$(basename $file .md)-compliance.json"
    
  # Flag high-risk content
  RISK_LEVEL=$(jq -r '.risk_assessment.level' "compliance-reports/$(basename $file .md)-compliance.json")
  if [[ "$RISK_LEVEL" == "high" ]]; then
    echo "High-risk content detected in $file" >> compliance-alerts.log
  fi
done

1
2

# Handle timeouts gracefully
timeout 300 claude -p "Long running analysis" || echo "Task timed out"

1
2
3
4
5
6
7
8

# Track costs across multiple operations
TOTAL_COST=0
for task in review test docs; do
  RESULT=$(claude -p "$task workflow" --output-format json)
  COST=$(echo "$RESULT" | jq '.cost.total')
  TOTAL_COST=$(echo "$TOTAL_COST + $COST" | bc)
done
echo "Total cost: $TOTAL_COST"

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16

# Robust error handling
run_claude_task() {
  local prompt="$1"
  local max_retries=3
  local retry=0
  
  while [ $retry -lt $max_retries ]; do
    if claude -p "$prompt" --output-format json > result.json 2>&1; then
      return 0
    fi
    retry=$((retry + 1))
    echo "Retry $retry/$max_retries after failure"
    sleep 5
  done
  return 1
}