Claude Continue

An iTerm2 daemon that monitors Claude Code sessions and automatically handles prompts, permissions, and continuations using smart regex patterns.

What It Does

When you run Claude Code in iTerm2, this daemon:

1. Auto-approves permission prompts - When Claude asks "Do you want to allow this operation?", automatically approves safe operations
2. Sends continuation commands - When Claude stops and asks "Continue?", automatically sends continue
3. Answers questions intelligently - Uses smart regex patterns to handle common prompts

Features

• Fast & Offline - No API calls, instant local responses
• Lightweight - Only requires iTerm2 Python API + aiohttp for web GUI
• Smart Patterns - Handles common Claude Code prompts intelligently
• Safe Defaults - Blocks dangerous operations (rm -rf, etc.)
• Web Control Panel - Monitor and control sessions via browser

Architecture

┌─────────────────────────────────────────────────────┐
│                    iTerm2                           │
│  ┌─────────────────────────────────────────────┐   │
│  │  Tab 1: Claude Code    Tab 2: Claude Code   │   │
│  └─────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────┘
                         │
                         │ Screen Streaming (iTerm2 Python API)
                         ▼
┌─────────────────────────────────────────────────────┐
│              SessionManager                          │
│  ┌─────────────────────────────────────────────┐   │
│  │  SessionMonitor 1   SessionMonitor 2   ...  │   │
│  └─────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────┘
                         │
                         ▼
┌─────────────────────────────────────────────────────┐
│              PatternDetector                         │
│  ┌─────────────────────────────────────────────┐   │
│  │  PERMISSION │ CONTINUATION │ QUESTION       │   │
│  └─────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────┘
                         │
          ┌──────────────┼──────────────┐
          ▼              ▼              ▼
     Auto "1"       Auto "continue"   Smart Regex
     (approve)                         Response

Installation

Requirements

• macOS
• iTerm2 (with Python API enabled)
• Python 3.12
• claudeContinue.sh requires a Python 3.12 venv named venv in the repo root

Quick Install

git clone https://github.com/jens-krypto/claude-continue.git
cd claude-continue
python3.12 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python src/daemon.py --setup

The interactive wizard will guide you through:

1. Behavior Mode - Full Auto, Semi Auto, Cautious, or Monitor Only
2. Timing Settings - Response speed configuration
3. Logging - Debug mode toggle
4. iTerm2 Installation - Optional AutoLaunch setup

Alternative: Script Install

./scripts/install.sh

Manual Install

1. Install dependencies:

pip install -r requirements.txt

2. Enable iTerm2 Python API:

   • Open iTerm2 Preferences (Cmd+,)
   • Go to "General" > "Magic"
   • Enable "Enable Python API"

3. Create AutoLaunch symlink:

mkdir -p ~/Library/Application\ Support/iTerm2/Scripts/AutoLaunch
ln -s $(pwd) ~/Library/Application\ Support/iTerm2/Scripts/AutoLaunch/claude-continue

4. Restart iTerm2

Usage

Automatic Mode (Recommended)

After installation, the daemon starts automatically when iTerm2 launches. It monitors all terminal sessions for Claude Code activity.

Manual Mode

# Run with iTerm2 integration
./scripts/run.sh

# Test without iTerm2 (pattern detection only)
./scripts/run.sh --test

# Debug mode
./scripts/run.sh --debug

# Re-run setup wizard
./scripts/run.sh --setup

# Reset config and re-run wizard
./scripts/run.sh --reset

# Run without web GUI
./scripts/run.sh --no-web

Web Control Panel

The daemon includes a web GUI for controlling sessions:

http://localhost:7777

Features

• Session Overview - See all active Claude sessions in iTerm2
• Toggle Automation - Enable/disable automation per session
• Live Settings - Change auto-approve, auto-continue, answer questions on the fly
• Action Tracking - See how many prompts have been handled per session

Disabling the Web GUI

python src/daemon.py --no-web

Configuration

Edit config/config.py or set environment variables:

Behavior

export CLAUDE_CONTINUE_AUTO_APPROVE="true"      # Auto-approve all permissions
export CLAUDE_CONTINUE_AUTO_CONTINUE="true"     # Auto-send continue
export CLAUDE_CONTINUE_ANSWER_QUESTIONS="true"  # Auto-answer questions
export CLAUDE_CONTINUE_COOLDOWN="1.0"           # Seconds between actions

Logging

export CLAUDE_CONTINUE_LOG_LEVEL="INFO"         # DEBUG, INFO, WARNING, ERROR
export CLAUDE_CONTINUE_LOG_FILE="~/Library/Logs/claude-continue.log"
export CLAUDE_CONTINUE_DEBUG="false"            # Extra verbose logging

macOS Permissions

iTerm2 Python API

The daemon requires iTerm2's Python API to be enabled:

1. Open iTerm2 Preferences (Cmd+,)
2. Go to General → Magic
3. Check Enable Python API
4. Restart iTerm2

Chrome Tab Reuse (Optional)

When the daemon starts, it opens the web GUI in your browser. To reuse an existing tab instead of opening a new one:

1. Open System Settings (or System Preferences on older macOS)
2. Go to Privacy & Security → Automation
3. Find iTerm in the list
4. Enable Google Chrome ✅

Without this permission, a new browser tab will open each time.

Accessibility (If Needed)

If you encounter issues with the daemon detecting or sending keystrokes:

1. Open System Settings → Privacy & Security → Accessibility
2. Add iTerm to the list and enable it

File Structure

claude-continue/
├── README.md                 # This file
├── LICENSE                   # GPL v3
├── requirements.txt          # Python dependencies
├── config/
│   ├── __init__.py
│   └── config.py             # Configuration settings
├── src/
│   ├── __init__.py
│   ├── daemon.py             # Main entry point
│   ├── session_monitor.py    # iTerm2 session monitoring
│   ├── pattern_detector.py   # Prompt pattern detection
│   ├── smart_responder.py    # Regex-based response logic
│   └── wizard.py             # Setup wizard
├── web/
│   ├── __init__.py
│   └── server.py             # Web GUI server (port 7777)
├── scripts/
│   ├── install.sh            # Install to iTerm2
│   ├── uninstall.sh          # Remove from iTerm2
│   └── run.sh                # Run manually
└── tests/
    ├── test_pattern_detector.py
    └── test_smart_responder.py

Security Risks & Considerations

╔═══════════════════════════════════════════════════════════════════════════╗
║  ⚠️  WARNING: This software automates command execution.                  ║
║      You run this service entirely at your own risk.                      ║
╚═══════════════════════════════════════════════════════════════════════════╝

Before enabling any automation level, please understand the implications:

Risk Overview

┌─────────────────────┬────────────────┬──────────────────────────────────────┐
│ Setting             │ Risk Level     │ What Could Go Wrong                  │
├─────────────────────┼────────────────┼──────────────────────────────────────┤
│ Auto-Approve        │ HIGH           │ Unintended file changes/deletions    │
│ Auto-Continue       │ MEDIUM         │ Operations continue without review   │
│ Answer Questions    │ HIGH           │ Wrong answers to important decisions │
│ Auto Follow-up      │ MEDIUM-HIGH    │ Unwanted actions in idle sessions    │
└─────────────────────┴────────────────┴──────────────────────────────────────┘

Auto-Approve Permissions

What it does: Automatically presses "1" (Yes) when Claude asks for permission to run commands.

Real risks:

• Claude might delete files you didn't want deleted
• Could modify code in unexpected ways
• May execute shell commands with unintended side effects
• Git operations (commits, pushes) happen without review

When to enable: Only when working on non-critical code where you're comfortable with Claude having full autonomy. Never use on production systems.

Recommendation: Start with this OFF. Review the patterns in smart_responder.py and the dangerous command blocklist before enabling.

Auto-Continue

What it does: Sends "continue" when Claude pauses or asks if you want it to proceed.

Real risks:

• Long refactoring operations continue without checkpoints
• May continue down a wrong path without your course correction
• Resource-intensive operations run to completion

When to enable: When you trust Claude's current direction and want uninterrupted work sessions.

Recommendation: Generally safer than Auto-Approve, but review Claude's plan before enabling.

Answer Questions

What it does: Uses regex patterns to automatically answer Claude's questions (like "Which option?" → "1").

Real risks:

• May choose wrong options for important decisions
• Could provide incorrect file names or paths
• Answers are based on simple pattern matching, not understanding

When to enable: Only for very routine tasks where the questions are predictable.

Recommendation: Keep OFF. This is disabled by default for good reason.

Auto Follow-up

What it does: Sends prompts to Claude when it appears idle (e.g., "What's next?").

Real risks:

• May trigger unwanted actions when you stepped away
• Could start new tasks you didn't intend
• May confuse Claude's context

When to enable: Only during active pair-programming sessions where you're monitoring.

Recommendation: Keep OFF unless actively working.

General Safety Tips

1. Start conservative - Begin with all settings OFF, enable one at a time
2. Monitor the logs - Check ~/Library/Logs/claude-continue.log regularly
3. Use per-session controls - Disable automation for sensitive sessions via the web GUI
4. Review the blocklist - Check src/smart_responder.py for blocked dangerous commands
5. Keep backups - Use git and don't auto-approve in repos without commits
6. Test first - Run with --test flag to see pattern matching without actions

What We Block

The daemon attempts to block obviously dangerous commands like:

• rm -rf / and similar destructive operations
• curl ... | bash (remote code execution)
• SQL DROP/DELETE statements
• Fork bombs and reverse shells
• Git force pushes

However, no blocklist is perfect. Novel dangerous commands may slip through.

---

Smart Responder Patterns

Safe Operations (Auto-Approved)

• Read operations: read file.py
• Edit code files: edit main.py, edit component.tsx
• Git status/diff/log
• Running tests: pytest, npm test

Dangerous Operations (Auto-Denied)

• rm -rf / or rm -rf ~
• curl ... | bash
• Editing .env files
• SQL DROP/DELETE statements

Question Responses

• "Would you like me to continue?" → continue
• "Do you want to create the file?" → yes
• "Which option?" → 1

Troubleshooting

Daemon not starting

1. Check if Python API is enabled in iTerm2 preferences
2. Check logs: tail -f ~/Library/Logs/claude-continue.log
3. Run manually: ./scripts/run.sh --debug

Not detecting prompts

1. Run test mode: ./scripts/run.sh --test
2. Check pattern detector output
3. Adjust patterns in src/pattern_detector.py

Uninstall

./scripts/uninstall.sh

About Anomaly Alpha

This project is brought to you by Anomaly Alpha, an AI-focused crypto project building tools and infrastructure for the intersection of artificial intelligence and blockchain technology.

We've open-sourced Claude Continue to share it with the developer community. If you find it useful, feel free to check out what else we're building.

• Website: addicted.bot
• X: @AddictedAnomaly
• Telegram: t.me/AnomalyAlpha

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add some amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

License

This project is licensed under the GPL v3 License - see the LICENSE file for details.