Thinking Modes

Control Claude Code's reasoning depth to match your task complexity. Four distinct modes optimize response quality and performance for different scenarios.

How to Activate Thinking Modes

Simply mention the mode in your request:

"Think hard about this authentication flow issue"
"Use ultra hard mode to analyze our database schema"  
"Standard thinking is fine for this CSS update"

Claude often infers the appropriate mode from context:

"This is critical - we need to be absolutely sure" → Higher mode
"Quick question about syntax" → Standard mode
"Deep dive into this performance issue" → Think Harder

You can also set default preferences in your CLAUDE.md:

# CLAUDE.md

## Thinking Preferences
- Use Think Hard for all debugging tasks
- Use Think Harder for architecture decisions
- Standard mode for routine development
- Ultra Hard for security-related code

What You Get with Each Mode

Standard Mode (Default)

Quick, direct responses for routine tasks:

"Add a loading spinner to the user dashboard"

Output: Fast implementation with standard patterns

• Response time: Seconds
• Depth: Surface-level
• Best for: Daily coding tasks, simple fixes, basic explanations

Think Hard Mode

Enhanced analysis for complex problems:

"Think hard about why our WebSocket connections drop after 5 minutes in production"

Output: Systematic investigation with multiple hypotheses

• Response time: 10-20 seconds
• Depth: Thorough consideration of edge cases
• Best for: Debugging, performance issues, integration problems

Think Harder Mode

Comprehensive analysis for architectural decisions:

"Think harder about implementing user permissions for multi-tenant SaaS with efficient API checking"

Output: Multiple solution comparison with trade-offs

• Response time: 20-40 seconds
• Depth: Deep analysis of implications
• Best for: System design, security analysis, complex refactoring

Ultra Hard Mode

Maximum reasoning for critical decisions:

"Ultra hard: Evaluate migrating from monolithic to microservices considering team size and growth projections"

Output: Exhaustive analysis with implementation roadmap

• Response time: 40+ seconds
• Depth: Long-term implications with detailed pros/cons
• Best for: Major architectural shifts, technology stack decisions

When to Use Each Mode

Standard Mode Selection

Use for routine development tasks where speed matters:

• Simple feature additions
• Code formatting and cleanup
• Basic explanations
• Quick syntax questions
• Routine refactoring

Think Hard Mode Selection

Use for problems requiring systematic investigation:

• Complex debugging and troubleshooting
• Performance optimization issues
• Integration problems
• Code reviews and analysis
• Understanding legacy code

Think Harder Mode Selection

Use for architectural decisions with long-term impact:

• System design and architecture
• Complex refactoring strategies
• Security analysis and implementation
• API design decisions
• Technology stack evaluations

Ultra Hard Mode Selection

Use for critical decisions affecting entire systems:

• Major architectural migrations
• Security-critical implementations
• Technology stack changes
• Disaster recovery planning
• Production system modifications

Common Patterns

Mode Escalation

Start simple, escalate when needed:

1. "Why is this test failing?"                    # Standard first
2. "Think hard about this failing test"           # If unclear
3. "Think harder - deep analysis of test failure" # If still stuck

Task-Mode Matching

Task Type	Mode	Why
CRUD operations	Standard	Straightforward patterns
Bug debugging	Think Hard	Systematic investigation
Performance tuning	Think Harder	Complex trade-offs
Security features	Ultra Hard	Critical to get right
API design	Think Harder	Long-term implications

Time-Boxed Thinking

Balance depth with constraints:

"Think hard for a quick analysis, we can go deeper later"
"Ultra hard mode but focus on key points"

Quick Reference

Activation Commands

"think hard"      # Enhanced analysis
"think harder"    # Deep analysis  
"ultra hard"      # Maximum reasoning
"standard"        # Return to default

Mode Selection Guide

Keywords	Mode
"quick", "simple", "basic"	Standard
"complex", "debug", "investigate"	Think Hard
"architecture", "design", "analyze"	Think Harder
"critical", "security", "migration"	Ultra Hard

Performance vs Quality

Mode	Response Time	Analysis Depth	Best For
Standard	Seconds	Surface	Daily tasks
Think Hard	10-20s	Thorough	Problem solving
Think Harder	20-40s	Deep	Architecture
Ultra Hard	40s+	Exhaustive	Critical decisions