Using AI coding assistants like Copilot, Cursor, or Claude Code? You might be:
- Shipping more bugs: AI-generated code has 1.7x more defects than human-written code (CodeRabbit 2025)
- Overestimating accuracy: Developers overestimate AI effectiveness by ~20% (METR Study)
- Slowing down: Junior developers take 7-10% longer with AI assistance (McKinsey)
- Losing skills: Blind approval of AI suggestions leads to "cognitive debt" and skill degradation
CodePause tracks your AI usage in real-time and provides skill-level aware guidance to help you:
- Maintain code ownership - Know when you're over-relying on AI
- Catch bugs earlier - Thoughtful review prevents blind approval
- Preserve critical thinking - Balance AI assistance with hands-on coding
- Build better habits - Gentle nudges, not intrusive warnings
CodePause runs silently in the background, tracking three key metrics:
- AI vs Manual Balance - Percentage of code generated by AI vs written by you
- Review Quality Score - How thoroughly you review AI suggestions (0-100)
- Balance Score - Overall health combining AI usage, review quality, and patterns
You get respectful, skill-level aware guidance when patterns emerge—not annoying popups.
Clean dashboard showing your three core metrics at a glance
CodePause focuses on what matters:
1. AI vs Manual Balance
- Track percentage of AI-generated vs manually-written code
- Skill-level targets: Junior <60%, Mid <50%, Senior <40%
- Trend analysis over 7-30-90 days
Track your progress with 7-day AI usage trends
2. Review Quality Score (0-100)
- Measures how thoroughly you review AI code
- Based on: time in focus (40%), scrolling (20%), cursor movement (20%), edits (20%)
- Categories: Thorough (70+), Light (40-69), Minimal (<40)
3. Balance Score
- Overall health indicator
- Combines AI/manual balance + review quality + patterns
- Ratings: Excellent, Good, Needs Attention
Different guidance for different experience levels:
Junior Developers:
- Educational tone with learning opportunities
- "Building foundational skills requires hands-on practice..."
- More frequent guidance (hourly)
Mid-Level Developers:
- Balanced, professional tone
- Data-driven recommendations
- Moderate frequency
Senior Developers:
- Evidence-based, research-cited
- Pattern analysis and insights
- Least intrusive (daily)
- Snooze for 24 hours - Take a break from notifications
- Dismiss permanently - Customize what you see
- Adaptive frequency - Low/medium/high settings
- Never intrusive - No popups during active typing
Detects AI code from ANY tool with 99.99% accuracy:
- Method 1: VS Code Inline Completion API (Copilot, Cursor)
- Method 2: Large paste detection (>100 characters)
- Method 3: External file changes (agent mode)
- Method 4: Git commit markers (Claude Code, etc.)
- Method 5: Change velocity analysis (typing speed)
No tool-specific configuration needed!
CodePause supports 15 programming languages with intelligent complexity-aware review time calculations:
Highest Complexity (requires most careful review):
- Rust (2.0x) - Memory safety, borrow checker
- C++ (1.8x) - Manual memory management, 70% of CVEs
- C (1.7x) - Buffer overflows, pointer arithmetic
- Scala (1.7x) - Advanced FP, steep learning curve
High Complexity:
- Java (1.6x) - Verbose, established patterns
- TypeScript (1.5x) - Structural typing, 15-19% bug detection
- JavaScript (1.5x) - Dynamic typing, inconsistent patterns
- C# (1.5x) - .NET complexity, unified type system
Moderate to High:
- Swift (1.4x), Kotlin (1.4x), Python (1.4x), Go (1.4x)
Moderate:
- Ruby (1.3x), PHP (1.2x)
What do these multipliers mean?
- A 2.0x multiplier means Rust code needs twice the review time compared to baseline
- Multipliers are based on empirical research: memory safety, bug density, learning curves, and cyclomatic complexity
- CodePause automatically adjusts expected review times based on the language you're working in
See full research and methodology: LANGUAGE_COMPLEXITY.md
No clutter. Just insights.
- 3 metric cards (AI%, Manual%, Balance)
- 7-day trend visualization
- Current session stats
- Quick actions (view details, adjust targets, snooze)
Removed: Gamification, XP levels, achievements, tool breakdowns
Seamlessly integrated into your VS Code workflow
Understand how you interact with AI - Agent Mode, Inline Autocomplete, and Chat/Paste
- Open VS Code
- Press
Ctrl+Shift+X(Windows/Linux) orCmd+Shift+X(Mac) - Search for "CodePause"
- Click Install
# Clone repository
git clone https://github.com/codepause-dev/codepause-extension.git
cd vscode-extension
# Install dependencies
npm install
# Compile TypeScript
npm run compile
# Run tests
npm test
# Launch Extension Development Host
# Press F5 in VS CodeFirst time? CodePause will ask:
Welcome to CodePause!
What's your experience level?
[I'm Junior] [I'm Mid-Level] [I'm Senior]
This adjusts notification frequency and thresholds.
Press Ctrl+Shift+P → "CodePause: Open Dashboard"
Or click the CodePause icon in the activity bar.
CodePause runs silently in the background, tracking:
- AI vs manual code balance
- Review quality for AI suggestions
- Patterns and trends
When needed, CodePause provides:
- Gentle reminders to review code
- Balance alerts if over-relying on AI
- Positive reinforcement when doing well
CodePause can be customized to fit your workflow. See Configuration Guide for:
- All available settings
- Experience level thresholds
- Command reference
Quick access: Ctrl+, → Search for "CodePause"
Have questions? Check our FAQ for answers to common questions about:
- Performance impact
- Data privacy
- Compatibility
- Customization
- Export options
- Open source licensing
CodePause tracks three core metrics to help you maintain coding skills while using AI effectively:
- AI Percentage - Your AI-to-manual code balance (Target: <50% for mid-level)
- Code Ownership Score - How thoroughly you review AI code (Target: 70+ for thorough review)
- Skill Development Health - Overall development health indicator (Excellent / Good / Needs Attention)
For detailed information on all metrics, see:
- Review Quality Metrics
- AI Usage Patterns (Agent Mode, Inline Autocomplete, Chat/Paste)
- Consistency Metrics (Streak Days, Unreviewed Files)
View the complete metrics documentation in the docs/metrics directory.
CodePause uses 5 detection methods:
-
Inline Completion API (High confidence)
- Official VS Code API
- Detects Copilot, Cursor inline suggestions
-
Large Paste Detection (High confidence)
- Single change >100 characters
- Analyzes code structure (braces, keywords)
-
External File Changes (High confidence)
- File modified while not open
- Indicates agent/autonomous mode
-
Git Commit Markers (Absolute confidence)
- Detects "Co-Authored-By: Claude", "@claude-code"
- Explicit AI markers
-
Change Velocity (Medium confidence)
-
500 characters in <1 second
- Pattern-based heuristic
-
Priority System: If conflict, highest confidence method wins.
reviewScore =
(timeInFocus × 40%) + // Did you spend time reviewing?
(scrollActivity × 20%) + // Did you read through code?
(cursorMovement × 20%) + // Did you navigate the code?
(editsMade × 20%) // Did you modify anything?
// Expected review time
expectedTime = linesOfCode × 500ms (min 10s, max 120s)
// Scoring
70-100: Thorough (green)
40-69: Light (yellow)
0-39: Minimal (needs attention)- Tracks coding sessions with 5-minute idle timeout
- Aggregates AI vs manual code per session
- Provides end-of-session summaries
- Privacy & Data - What we track and how your data is stored
- Research & Evidence - Studies backing CodePause's approach
- Roadmap - Future features and improvements
- Known Issues - Current limitations and workarounds
We welcome contributions!
- Fork the repository
- Create feature branch:
git checkout -b feature/amazing-feature - Make changes and write tests
- Commit:
git commit -m "Add amazing feature" - Push:
git push origin feature/amazing-feature - Open Pull Request
- Follow existing code style (TypeScript + ESLint)
- Write tests for new features (Jest)
- Update documentation
- Keep commits focused
- Be respectful and constructive
- Bug fixes and testing
- Documentation improvements
- Translations
- UI/UX enhancements
- Research on effectiveness
Business Source License 1.1
Allowed:
- Personal use (free)
- Internal company use (free)
- Modify and contribute
- Learn from the code
Not Allowed:
- Creating competing commercial extension
- Selling this software
- Offering as SaaS
Future: Becomes Apache 2.0 on January 4, 2027
Commercial inquiries: license@codepause.dev
- Documentation: docs.codepause.dev
- Discussions: GitHub Discussions
- Bug Reports: GitHub Issues
- Email: support@codepause.dev
Built with:
- VS Code Extension API
- TypeScript
- SQLite (sql.js)
- Jest (testing)
Inspired by research on AI pair programming effectiveness and developer productivity.
CodePause is designed to help developers maintain code ownership and quality. It is:
- NOT a substitute for proper code review
- NOT a guarantee against bugs
- NOT affiliated with GitHub, Anthropic, or Cursor
- NOT collecting or selling your data
Use responsibly. Developers not liable for issues arising from use.
Made with care by developers, for developers
Star on GitHub • Report Bug • Request Feature
Pause. Review. Own your code.