Zellij MCP Server
Enables comprehensive management of Zellij terminal workspace sessions, including session, pane, tab, plugin, layout operations, and LLM completion detection.
README
Zellij MCP Server
A comprehensive Model Context Protocol (MCP) server for managing Zellij terminal workspace sessions. This server provides extensive tools for session management, pane operations, tab controls, plugin integration, layout management, and advanced LLM completion detection.
Features
🏗️ Session Management
Complete session lifecycle management with advanced operations:
- Basic Operations: List, create, attach, kill, delete sessions
- Advanced Operations: Clone, rename, switch, export/import session configurations
- Session Information: Detailed session info and health monitoring
🔧 Pane Management
Comprehensive pane control with advanced positioning:
- Basic Operations: Create, close, focus, resize, swap panes
- Advanced Features: Stack panes, floating pane controls, fullscreen toggle
- Content Operations: Clear, dump screen, edit scrollback, scroll controls
- Positioning: Precise floating coordinate control
📑 Tab Management
Full tab lifecycle with navigation and organization:
- Tab Operations: Create, close, rename, move tabs
- Navigation: Go to tab by index/name, next/previous navigation
- Advanced Features: Query tab names, synchronized input across panes
🔌 Plugin Management
Comprehensive plugin ecosystem integration:
- Plugin Lifecycle: Launch, focus, start/reload plugins
- Advanced Options: Floating, in-place, skip cache configurations
- Plugin Information: List aliases, running plugins, plugin info
- Configuration: Full plugin configuration support
📐 Layout Management
Complete layout system with validation:
- Layout Operations: Dump, save, apply, load layouts
- Advanced Features: New tab with layout, layout validation
- Layout Discovery: List available layouts with custom directory support
🔄 Piping System
Advanced inter-plugin communication:
- Basic Piping: Send data to plugins, broadcast to all plugins
- Advanced Piping: Response capture, file-based piping
- Action Piping: Force launch, skip cache, floating options
- Plugin Targeting: Specific plugin URL targeting
🤖 LLM Completion Detection
Specialized tools for LLM workflow integration:
- Process Monitoring: Watch pipes, poll processes, file watching
- Timeout Management: Automatic timeout completion
- Wrapper Scripts: LLM completion detector wrappers
- Resource Cleanup: Automated cleanup of detection resources
🛡️ Security & Performance
Enterprise-grade security and performance features:
- Input Validation: Comprehensive input sanitization and validation
- Rate Limiting: Configurable rate limiting per tool
- Caching: Intelligent caching with statistics
- Health Monitoring: System health checks and diagnostics
Installation
Prerequisites
- Node.js 18 or higher
- Zellij installed and available in PATH
- Active terminal environment
Setup
# Clone the repository
git clone https://github.com/yourusername/zellij-mcp-server.git
cd zellij-mcp-server
# Install dependencies
npm install
# Build the project
npm run build
# Make the script executable
chmod +x dist/index.js
Usage
With Claude Desktop
Add this to your Claude Desktop configuration file:
{
"mcpServers": {
"zellij": {
"command": "node",
"args": ["/path/to/zellij-mcp-server/dist/index.js"]
}
}
}
Direct Usage
# Start the MCP server
npm start
# Development mode with watching
npm run dev
Tool Categories
Session Management Tools
| Tool | Description |
|---|---|
zellij_list_sessions |
List all active Zellij sessions with caching |
zellij_new_session |
Create a new session with optional layout |
zellij_attach_session |
Attach to an existing session |
zellij_kill_session |
Kill a specific session |
zellij_delete_session |
Delete a specific session |
zellij_get_session_info |
Get detailed session information |
zellij_export_session |
Export session configuration to JSON |
zellij_import_session |
Import session from JSON export |
zellij_clone_session |
Clone an existing session |
zellij_rename_session |
Rename a session |
zellij_switch_session |
Switch to a different session |
Pane Management Tools
| Tool | Description |
|---|---|
zellij_new_pane |
Create a new pane with advanced options |
zellij_close_pane |
Close the currently focused pane |
zellij_focus_pane |
Focus pane in specific direction |
zellij_resize_pane |
Resize the focused pane |
zellij_swap_panes |
Move/swap pane in specified direction |
zellij_stack_panes |
Stack multiple panes by their IDs |
zellij_toggle_floating |
Toggle floating panes visibility |
zellij_toggle_fullscreen |
Toggle fullscreen mode for focused pane |
zellij_pin_pane |
Pin/unpin floating pane |
zellij_clear_pane |
Clear the focused pane buffer |
zellij_dump_screen |
Dump pane screen content to file |
zellij_edit_scrollback |
Edit pane scrollback in default editor |
zellij_rename_pane |
Rename the focused pane |
zellij_scroll |
Scroll in focused pane |
zellij_exec_in_pane |
Execute command in current pane |
zellij_write_to_pane |
Write text to current pane |
zellij_get_pane_info |
Get current pane layout information |
Tab Management Tools
| Tool | Description |
|---|---|
zellij_new_tab |
Create a new tab |
zellij_close_tab |
Close the current tab |
zellij_rename_tab |
Rename the current tab |
zellij_go_to_tab |
Go to tab by index |
zellij_go_to_tab_name |
Go to tab by name |
zellij_move_tab |
Move tab left or right |
zellij_query_tab_names |
Get all tab names in current session |
zellij_toggle_sync_tab |
Toggle synchronized input across panes |
zellij_go_to_next_tab |
Switch to next tab |
zellij_go_to_previous_tab |
Switch to previous tab |
Plugin Management Tools
| Tool | Description |
|---|---|
zellij_launch_plugin |
Launch a plugin with full configuration |
zellij_action_launch_plugin |
Launch plugin using action command |
zellij_launch_or_focus_plugin |
Smart plugin activation |
zellij_start_or_reload_plugin |
Start or reload a plugin |
zellij_list_aliases |
List available plugin aliases |
zellij_get_plugin_info |
Get information about a plugin |
zellij_list_running_plugins |
List currently running plugins |
Layout Management Tools
| Tool | Description |
|---|---|
zellij_dump_layout |
Dump current layout to stdout or file |
zellij_save_layout |
Save current layout to layouts directory |
zellij_apply_layout |
Apply a layout to current or new session |
zellij_list_layouts |
List available layouts |
zellij_load_layout |
Load and display layout content |
zellij_new_tab_with_layout |
Create new tab with specific layout |
zellij_validate_layout |
Validate layout file syntax |
Piping System Tools
| Tool | Description |
|---|---|
zellij_pipe |
Send data to plugins via pipe with advanced options |
zellij_pipe_to_plugin |
Send data to a specific plugin |
zellij_pipe_broadcast |
Broadcast data to all listening plugins |
zellij_action_pipe |
Advanced piping with action-level options |
zellij_pipe_with_response |
Send data and capture plugin response |
zellij_pipe_from_file |
Pipe file content to plugins |
LLM Detection Tools
| Tool | Description |
|---|---|
zellij_watch_pipe |
Watch a pipe for specific patterns with timeout |
zellij_create_named_pipe |
Create a named pipe for bidirectional communication |
zellij_pipe_with_timeout |
Pipe command output with automatic timeout |
zellij_poll_process |
Poll process status by PID |
zellij_watch_file |
Watch file for changes with pattern matching |
zellij_create_llm_wrapper |
Create LLM completion detector wrapper script |
zellij_cleanup_detection |
Clean up detection resources |
System Tools
| Tool | Description |
|---|---|
zellij_run_command |
Run command in new pane |
zellij_edit_file |
Edit file in new Zellij pane |
zellij_switch_mode |
Switch input mode for all connected clients |
zellij_kill_all_sessions |
Kill all active sessions |
zellij_delete_all_sessions |
Delete all sessions |
zellij_clear_cache |
Clear MCP server cache |
zellij_get_cache_stats |
Get cache statistics |
zellij_health_check |
Perform system health check |
Example Usage
Creating a Development Session
{
"name": "zellij_new_session",
"arguments": {
"session_name": "development",
"layout": "dev-layout"
}
}
Running Commands in Panes
{
"name": "zellij_run_command",
"arguments": {
"command": "npm run dev",
"direction": "right"
}
}
Plugin Management
{
"name": "zellij_launch_plugin",
"arguments": {
"plugin_url": "file:~/.config/zellij/plugins/filepicker.wasm",
"floating": true,
"width": "50%",
"height": "50%"
}
}
Layout Operations
{
"name": "zellij_save_layout",
"arguments": {
"layout_name": "my-dev-setup"
}
}
LLM Integration
Creating LLM Wrapper Scripts
The zellij_create_llm_wrapper tool generates intelligent wrapper scripts that provide robust completion detection for any LLM command:
{
"name": "zellij_create_llm_wrapper",
"arguments": {
"wrapper_name": "claude-wrapper",
"llm_command": "claude chat",
"detect_marker": "<<<COMPLETE>>>",
"timeout_ms": 60000
}
}
What the wrapper provides:
- Multi-signal detection: Exit codes, completion markers, and status files
- Automatic timeout handling: Configurable timeout with graceful cleanup
- Process monitoring: Real-time status tracking and logging
- Signal handling: Proper cleanup on interruption (SIGINT, SIGTERM)
- Timestamped logging: Detailed execution logs for debugging
Generated files:
/tmp/llm-wrapper-{name}.sh- The executable wrapper script/tmp/llm-status-{name}- Real-time status file with timestamps/tmp/llm-output-{name}-{pid}- Captured output (temporary)
Usage example:
# After creating the wrapper, use it like:
/tmp/llm-wrapper-claude-wrapper.sh "Explain quantum computing"
# Monitor status in real-time:
tail -f /tmp/llm-status-claude-wrapper
# The wrapper handles timeouts, signals, and cleanup automatically
Status tracking:
running- LLM query is in progresscomplete:0- Successfully completedtimeout- Query timed outerror:N- Failed with exit code N
Architecture
Security Features
- Input Validation: All inputs are validated and sanitized
- Rate Limiting: Per-tool rate limiting (50 requests per minute)
- Path Traversal Protection: File path validation prevents directory traversal
- Command Injection Prevention: Command sanitization and validation
Performance Features
- Intelligent Caching: Session lists and other data are cached
- Async Operations: All operations are asynchronous
- Error Handling: Comprehensive error handling with specific error types
- Resource Cleanup: Automatic cleanup of temporary resources
Error Handling
The server includes comprehensive error handling with specific error types:
ValidationError: Input validation failuresSecurityError: Security-related errors (rate limiting, etc.)ZellijError: Zellij-specific command failures
Development
Project Structure
src/
├── index.ts # Main MCP server implementation
├── tools/ # Tool modules
│ ├── detection.ts # LLM completion detection tools
│ ├── layouts.ts # Layout management tools
│ ├── panes.ts # Pane management tools
│ ├── piping.ts # Piping system tools
│ ├── plugins.ts # Plugin management tools
│ └── sessions.ts # Session management tools
├── types/
│ └── zellij.ts # TypeScript type definitions
└── utils/
├── cache.ts # Caching utilities
├── command.ts # Command execution utilities
└── validator.ts # Input validation utilities
Building
# Development build with watching
npm run dev
# Production build
npm run build
# Start the built server
npm start
Testing
Run ./test-workflow.sh to validate the detection system functionality.
Contributing
- Fork the repository
- Create a feature branch (
git checkout -b feature/amazing-feature) - Commit your changes (
git commit -m 'Add some amazing feature') - Push to the branch (
git push origin feature/amazing-feature) - Open a Pull Request
License
This project is licensed under the MIT License - see the LICENSE file for details.
Acknowledgments
- Zellij - The amazing terminal workspace manager
- Model Context Protocol - The protocol this server implements
- Claude Code - Integration with Claude AI
Support
For issues, questions, or contributions, please visit the GitHub repository.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。