JSPM

  • Created
  • Published
  • Downloads 44
  • Score
    100M100P100Q78388F
  • License MIT

Run Claude Code, Gemini, Codex — or any coding agent — in a clean, isolated sandbox with sensitive data redaction and observability baked in.

Package Exports

  • vibekit
  • vibekit/dist/cli.js

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (vibekit) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

Vibekit CLI

A CLI middleware for headless and TUI coding agents that provides logging and sandbox capabilities.

Overview

Vibekit CLI wraps existing coding agents like Claude Code CLI, Gemini CLI, and others to provide:

  • Docker Security: Agents run in isolated containers with no host system access
  • Logging: Capture all agent interactions and commands
  • Safe Sandboxing: Complete isolation with resource limits and network restrictions
  • Unified Interface: Single CLI for multiple coding agents

Installation

npm install -g vibekit-cli

Or install locally:

npm install
npm link

Usage

Basic Commands

# Run Claude with local sandbox (default, no dependencies required)
vibekit claude

# Run with different sandbox types
vibekit claude --sandbox local        # File isolation (default, fast)
vibekit claude --sandbox docker       # Container isolation (requires Docker)
vibekit claude --sandbox none         # No sandbox (dangerous!)

# Allow network access
vibekit claude --network              # Enable network access

# Pass arguments to the underlying agent  
vibekit claude --help
vibekit claude "Fix the bug in src/app.js"

# View logs
vibekit logs
vibekit logs --agent claude --lines 100

# Open analytics dashboard
vibekit dashboard                     # Start dashboard and open in browser
vibekit status                        # View dashboard status

# Sync changes from sandbox back to your project
vibekit sync

# Clean up
vibekit clean
vibekit clean --logs
vibekit clean --docker

Sandbox Options

Local (Default - Fast & Simple):

  • Instant startup - No dependencies required
  • 📁 File isolation - Protects your main project files
  • 💻 Full terminal UI support - Colors, TUI, interactive prompts
  • 🏃 Fast execution - No container overhead
  • 🔄 Easy sync - Changes isolated until you decide to apply them
  • Works everywhere - No Docker installation needed

Docker (Maximum Security):

  • 🔒 Complete container isolation - Agents can't access host system
  • 💻 Full terminal UI support - Colors, TUI, interactive prompts
  • 🚫 No host system access - True containerization
  • 📊 Resource limits - Configurable CPU/memory limits
  • ⚙️ Requires Docker - Must have Docker installed

None (Dangerous):

  • No overhead - Direct execution
  • ⚠️ No protection - Agent has full system access
  • 🚨 Only use with trusted agents

Logging

All agent interactions are logged to ~/.vibekit/logs/:

  • Commands executed
  • Agent responses
  • Errors and debugging info
  • Execution time and metadata

Enable debug logging:

VIBEKIT_DEBUG=1 vibekit claude

Analytics Dashboard

Vibekit includes a built-in analytics dashboard that provides real-time insights into your coding sessions:

# Start dashboard and open in browser
vibekit dashboard

# Check dashboard status
vibekit status

Dashboard Features:

  • 📊 Session Analytics - Track coding session duration and productivity
  • 🔍 Command History - View all executed commands and their outcomes
  • Real-time Updates - Live monitoring of agent activity
  • 📈 Performance Metrics - Analyze agent response times and success rates
  • 🎯 Project Insights - Understand file modification patterns

The dashboard runs on http://localhost:3001 by default and provides a clean, modern interface for monitoring your AI-assisted coding workflows.

Configuration

Config file location: ~/.vibekit/config.json

{
  "agents": {
    "claude": {
      "command": "claude", 
      "args": [],
      "env": {},
      "sandbox": {
        "enabled": true,
        "autoBackup": true,
        "autoSync": true
      }
    }
  },
  "logging": {
    "level": "info",
    "debug": false,
    "retention": {
      "days": 30,
      "maxFiles": 100
    }
  },
  "sandbox": {
    "defaultEnabled": true,
    "backupOnStart": true,
    "syncOnExit": true
  }
}

Supported Agents

  • Claude Code CLI: vibekit claude
  • Gemini CLI: vibekit gemini
  • More agents can be easily added

Development

git clone <repo>
cd vibekit-cli
npm install
npm link

Architecture

vibekit-cli/
├── src/
│   ├── cli.js              # Main CLI entry point
│   ├── agents/             # Agent wrapper modules
│   │   ├── base.js         # Base agent class
│   │   ├── claude.js       # Claude Code CLI wrapper
│   │   └── gemini.js       # Gemini CLI wrapper
│   ├── logging/            # Logging system
│   ├── sandbox/            # Sandbox functionality
│   └── config/             # Configuration management
└── bin/vibekit             # Executable