Claude Multi-Terminal

Note for Terminal Agents: If you are a Claude instance running inside this application, you don't need to read this file. Please refer to CLAUDE.md for your instructions.

A multi-terminal editor for coordinating multiple Claude Code instances. Run multiple Claude sessions side-by-side with a unified chat interface to send commands to selected terminals.

Features

• Multi-Terminal Support: Run multiple Claude Code instances in a 2x2 grid layout
• Unified Chat Interface: Send commands to terminals using @ syntax
• Multi-@ Terminal Commands: Target terminals with flexible syntax:
  • @all command - Send to all terminals
  • @term1,term2,term3 command - Comma-separated terminals
  • @term1 @term2 command - Space-separated terminals
• Mixed @/ Commands: Combine terminal targeting with slash commands:
  • @term1 /clear - Run slash command on specific terminal
  • @term1 @term2 /list - Multiple terminals with slash commands
  • @all /interrupt - Slash commands with @all
• Smart Autocomplete: Get suggestions while typing @ or / anywhere in your message
• Slash Commands: Built-in commands for terminal management (works anywhere)
• Default Prompt: Set a default prompt to prepend to all messages
• Plan Mode Support: Automatically enters plan mode with Shift+Tab
• Auto-Start Claude: Each terminal automatically starts Claude Code
• Inter-Claude Communication: Claude instances can chat and coordinate using echo commands
• Terminal Management: Claude can create, close, and rename terminals autonomously
• Rename Terminals: Double-click terminal titles to rename them
• Responsive Layout: Adapts to different screen sizes
• Multiple Views: Messages, Terminals, Projects, and Board tabs for different workflows

What's New in v3.0

• Project Management Interface - Built-in epic creation and team collaboration tools
• Scrum Board View - Visual task tracking with swimlanes for each agent
• Enhanced Terminal Instructions - Terminals now start with "standby and wait for instructions"
• Bash Script Integration - Simplified communication using ./bin/send, ./bin/broadcast commands
• Agent Inbox System - Messages delivered to agent-specific inboxes with manual checking
• File System Integration - Project Work directory structure for epics, stories, and tasks

Breaking Changes in v2.2

• JSON-based IPC System - All Claude communication now uses JSON format
• File-based messaging - Claude writes all responses to .ipc-messages/ as JSON files
• No terminal output - Claude no longer outputs responses directly to terminal

Breaking Changes in v2.0

• Removed checkbox selection system - All commands must now use @ syntax
• No default terminal selection - You must explicitly specify terminals with @
• Plain text without @ or / will show an error - Every command needs a target

To migrate from v1.x:

• Instead of selecting checkboxes and typing a command, use: @term1 @term2 command
• To send to all terminals, use: @all command

Installation

From npm

npm install -g @adiontaegerron/claude-multi-terminal

From GitHub

git clone https://github.com/adiontae-tp/multi-terminal.git
cd multi-terminal
npm install
npm link

Usage

Navigate to your project directory and run:

claude-multi

This will open the multi-terminal editor with all terminals starting in your current directory.

Basic Workflow

1. Start the Application: Run claude-multi in your project directory
2. Create Terminals: Click "New Terminal" to add more Claude instances (starts with one)
3. Target Terminals: Use @ syntax to specify which terminals receive commands
4. Send Commands: Type your command with @ targeting and press Enter
5. Use Default Prompt: Expand the "Default Prompt" section to set a prompt that's prepended to all messages

Important: Version 2.0 removes checkbox selection. All commands must now use @ syntax to target terminals.

Slash Commands

• /send <terminal> <command> - Send command to specific terminal
• /send-all <command> or /broadcast <command> - Send to all terminals
• /list - List all terminals
• /create [name] - Create new terminal with optional name
• /close <terminal> - Close specific terminal
• /return <terminal> - Submit current input in terminal (press Enter)
• /clear <terminal> or /interrupt <terminal> - Send Ctrl+C to terminal
• /help - Show available commands

@ Terminal Commands

Send commands directly to terminals using @ syntax:

• @Terminal1 ls -la - Send to single terminal
• @"My Terminal" pwd - Terminal with spaces in name
• @term1,term2,term3 date - Multiple terminals (comma-separated)
• @term1 @term2 @term3 whoami - Multiple terminals (space-separated)
• @all echo "Hello" - Send to all terminals

Autocomplete: Type @ anywhere to get terminal suggestions!

Application Views

The application features four main views accessible via bottom tabs:

1. Messages Tab: View all inter-agent communication and system messages
2. Terminals Tab: Traditional terminal interface with multiple Claude instances
3. Projects Tab: Create epics, manage stories, and coordinate team tasks
4. Board Tab: Scrum board visualization with task swimlanes by agent

Claude-to-Claude Communication

Claude instances communicate by creating JSON files in the .ipc-messages/ directory. All communication uses a unified JSON format:

IPC Commands (trigger actions and display in Messages tab):

• Send message: {"type":"ipc","command":"SEND","from":"Terminal1","to":"Terminal2","content":"Hello"}
• Broadcast: {"type":"ipc","command":"BROADCAST","from":"Terminal1","content":"Hello all"}
• Create terminal: {"type":"ipc","command":"CREATE","from":"Terminal1","content":"NewTerminalName"}
• Close terminal: {"type":"ipc","command":"CLOSE","from":"Terminal1","to":"Terminal2"}
• Rename terminal: {"type":"ipc","command":"RENAME","from":"Terminal1","to":"Terminal2","content":"NewName"}
• List terminals: {"type":"ipc","command":"LIST","from":"Terminal1"}
• Get status: {"type":"ipc","command":"STATUS","from":"Terminal1","to":"Terminal2"}

Display Messages (UI only, no actions):

• User responses: {"type":"display","from":"Terminal1","content":"Task completed","metadata":{"category":"user","visible":true}}
• Status updates: {"type":"display","from":"Terminal1","content":"Processing...","metadata":{"category":"info","visible":true}}

All Claude responses appear in the Messages tab with clean formatting, icons, and timestamps.

Tips

• Each terminal automatically starts Claude Code when created
• Claude receives its identity: "You are a terminal agent. Your name is [Terminal Name]."
• Plan mode is activated automatically using Shift+Tab
• Double-click terminal titles to rename them for better organization
• The application uses your current directory as the working directory for all terminals
• Use multi-@ commands to coordinate multiple Claude instances efficiently
• Claude instances communicate instantly using IPC echo commands for seamless teamwork

Development

Prerequisites

• Node.js (v14 or higher)
• npm

Local Development

# Clone the repository
git clone https://github.com/adiontae-tp/multi-terminal.git
cd multi-terminal

# Install dependencies
npm install

# Start the application
npm start

Building

# Rebuild native modules
npm run rebuild

Requirements

• Claude Code: You must have Claude Code installed and accessible via the claude command
• Operating System: Works on macOS, Windows, and Linux
• Node.js: Requires Node.js 14 or higher

License

MIT

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/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

Issues

If you find a bug or have a feature request, please open an issue on GitHub.

Acknowledgments

• Built with Electron
• Terminal emulation by xterm.js
• PTY support from node-pty