🖥️ VirtualBox MCP Server

A powerful Model Context Protocol (MCP) server for managing VirtualBox VMs via Vagrant.

AI agents can now provision, manage, and debug virtual development environments with full observability.

Features • Quick Start • Tools • Workflows • Examples • Configuration

✨ Features

• 46 MCP Tools for complete VM lifecycle management
• Real-time Observability with logs, dashboards, and progress tracking
• Snapshot Management for safe rollback and recovery
• Process Control with kill/list capabilities
• File Synchronization with conflict resolution
• Async Operations with progress tracking and cancellation
• System Guardrails for zombie VM detection and cleanup
• Sequential Thinking for AI problem-solving

📦 Architecture

Virtualbox-mcp-server/          # Turborepo Monorepo
├── apps/
│   └── mcp-server/             # Main MCP server (46 tools)
│       └── src/
│           ├── index.ts        # Tool definitions & handlers
│           ├── error-handler.ts
│           ├── port-manager.ts
│           └── sequential-thinking.ts
├── packages/
│   ├── vagrant-client/         # Vagrant CLI wrapper
│   ├── sync-engine/            # Chokidar + file sync
│   └── shared-utils/           # Logger utilities
├── turbo.json
└── package.json

🚀 Quick Start

Prerequisites

• Node.js 18+
• VirtualBox 6.x or 7.x
• Vagrant 2.3+

Installation

# Install via NPM (Recommended)
npm install -g @use.manus.ai/virtualbox-mcp-server

# Or Clone and Build from Source
git clone https://github.com/usemanusai/Virtualbox-mcp-server.git
cd Virtualbox-mcp-server

# Install dependencies
npm install

# Build all packages
npm run build

Running the Server

# Run the installed server
virtualbox-mcp-server

# Or run from source
node apps/mcp-server/dist/index.js

🛠️ All 46 Tools

VM Lifecycle (9 tools)

| Tool | Description | |------|-------------| | create_vm | Create a new Vagrant VM | | create_dev_vm | Create VM with full config (CPU, memory, ports, sync) | | ensure_dev_vm | Start or create VM if not exists | | get_vm_status | Get VM state | | list_vms | List all VMs | | destroy_vm | Destroy VM (force) | | resize_vm_resources | Modify CPU/RAM/GUI settings | | package_box | Export VM as a .box file | | set_display_mode | Toggle Headless/GUI mode |

Execution (4 tools)

| Tool | Description | |------|-------------| | exec_command | Execute command in VM (with timeout) | | exec_with_sync | Execute with rsync before/after | | run_background_task | Run nohup background task | | atomic_transaction_exec | Execute with auto-rollback on failure |

Environment Setup (4 tools)

| Tool | Description | |------|-------------| | setup_dev_environment | Install runtimes (node, python, go, etc.) | | install_dev_tools | Install tools (git, docker, nginx, etc.) | | configure_shell | Configure aliases and env vars | | inject_secrets | Securely inject environment variables |

File Operations (7 tools)

| Tool | Description | |------|-------------| | upload_file | Upload file to VM | | search_files | Grep search in VM | | configure_sync | Configure file watcher | | sync_to_vm | Rsync host→VM | | sync_from_vm | Rsync VM→host | | sync_status | Get sync state | | resolve_conflict | Resolve sync conflicts |

👁️ Observability (3 tools)

| Tool | Description | |------|-------------| | tail_vm_log | Read last N lines of a log file (e.g., /var/log/syslog) | | get_task_output | Get stdout/stderr of background tasks | | grep_log_stream | Search for patterns in log files |

📸 Snapshots (4 tools)

| Tool | Description | |------|-------------| | snapshot_save | Create named snapshot before risky operations | | snapshot_restore | Revert to a specific snapshot | | snapshot_list | List all available snapshots | | snapshot_delete | Delete a specific snapshot |

⚙️ Process Control (2 tools)

| Tool | Description | |------|-------------| | list_processes | Return structured list of running processes (ps aux) | | kill_process | Send SIGTERM/SIGKILL to a process |

🌐 Network (1 tool)

| Tool | Description | |------|-------------| | check_vm_port | Verify if port is listening in VM & accessible from host |

📊 Dashboard (1 tool)

| Tool | Description | |------|-------------| | get_vm_dashboard | Comprehensive dashboard: CPU, RAM, Disk, tasks, logs |

⏳ Progress Awareness (6 tools)

| Tool | Description | |------|-------------| | start_download | Start tracked download, returns operation_id | | get_operation_progress | Get real-time progress (bytes, %, ETA) | | wait_for_operation | Block until operation completes or times out | | cancel_operation | Cancel a running operation | | list_active_operations | List all active operations | | sentinel_await | Wait for file, port, or service condition |

🛡️ Guardrails & Security (4 tools)

| Tool | Description | |------|-------------| | scan_system_health | Check disk/memory, identify Zombie VMs | | cleanup_zombies | Safely destroy orphaned VMs (with dry-run option) | | audit_security | Scan for open ports and weak configs | | forensic_blackbox_capture | Capture process & network state |

🧠 AI Reasoning (1 tool)

| Tool | Description | |------|-------------| | sequentialthinking | Dynamic problem-solving with reflection & branching |

🔄 Architectural Workflows

🟢 Easy Workflows

1. The "Daily Standup" (Environment Prep)

Quickly bring a dev environment online and ensure it's ready:

Workflow: ensure_dev_vm → sync_to_vm → install_dev_tools → get_vm_dashboard

1. ensure_dev_vm — Boot or create the VM automatically
2. sync_to_vm — Push local code changes to VM
3. install_dev_tools — Verify tools are present
4. get_vm_dashboard — Confirm VM is healthy

2. The "Dataset Fetch" (Async Download)

Download large files without blocking:

Workflow: start_download → wait_for_operation → search_files

1. start_download — Initiate download, get operation_id
2. wait_for_operation — Block until download completes
3. search_files — Verify file exists at expected path

3. The "Service Pulse" (Basic Debugging)

Quickly diagnose why localhost:8080 isn't loading:

Workflow: list_vms → check_vm_port → tail_vm_log

1. list_vms — Identify which VM handles the service
2. check_vm_port — Check if app is listening (vs. port forwarding issue)
3. tail_vm_log — Pull last 50 lines of error log

🔴 Advanced Workflows

4. The "Safety First" Update (Transactional Rollback)

Apply risky updates with a safety net:

Workflow: snapshot_save → start_download → check_vm_port → (rollback or delete snapshot)

1. snapshot_save — Create checkpoint "pre-update-v2"
2. start_download — Download new binary/patch
3. wait_for_operation — Block until complete
4. exec_command — Run installation script
5. check_vm_port — Verify service is back online
   • IF FAILED: snapshot_restore to rollback
   • IF SUCCESS: snapshot_delete to clean up

5. The "Resource Reclamation" (System Hygiene)

Identify and clean up orphaned Zombie VMs:

Workflow: scan_system_health → sequentialthinking → cleanup_zombies → get_vm_dashboard

1. scan_system_health — Identify Zombie VMs consuming resources
2. sequentialthinking — Analyze which are safe to delete
3. cleanup_zombies — Safely terminate with dry_run first
4. get_vm_dashboard — Confirm resources are freed

6. The "Deep Fix" Loop (Intelligent Debugging)

Autonomous diagnosis and repair of stuck processes:

Workflow: get_vm_dashboard → list_processes → grep_log_stream → sequentialthinking → kill_process → exec_with_sync

1. get_vm_dashboard — Detect CPU spike or stuck task
2. list_processes — Find the specific PID causing issues
3. grep_log_stream — Search logs for error signature
4. sequentialthinking — Formulate hypothesis
5. kill_process — Send SIGTERM to stuck process
6. exec_with_sync — Upload patched config and restart

💬 Example Prompts

Here are 30 natural language prompts with their corresponding tool calls:

1. Provision a New Database Server

"I need a fresh Redis server. Create a VM named 'redis-cache' using the 'hashicorp/bionic64' box."

{
  "name": "create_vm",
  "arguments": {
    "name": "redis-cache",
    "box": "hashicorp/bionic64"
  }
}

2. Status Check

"Is the 'frontend-react' VM currently running?"

{
  "name": "get_vm_status",
  "arguments": {
    "name": "frontend-react"
  }
}

3. Inventory Overview

"Show me a list of all the virtual machines we are currently managing."

{
  "name": "list_vms",
  "arguments": {}
}

4. Force Termination

"The 'experiment-01' VM is completely unresponsive. Destroy it immediately."

{
  "name": "destroy_vm",
  "arguments": {
    "name": "experiment-01"
  }
}

5. Idempotent Environment Check

"Ensure the 'integration-test' VM is running. If it's not there, create it."

{
  "name": "ensure_dev_vm",
  "arguments": {
    "name": "integration-test"
  }
}

6. Install Dependencies

"Run pip install -r requirements.txt inside the 'api-server' VM."

{
  "name": "exec_command",
  "arguments": {
    "vm_name": "api-server",
    "command": "pip install -r requirements.txt"
  }
}

7. Build and Sync

"Sync the latest changes to 'builder' and then run make build immediately."

{
  "name": "exec_with_sync",
  "arguments": {
    "vm_name": "builder",
    "command": "make build"
  }
}

8. Long-Running Job

"Start the data ingestion script (python ingest.py) on 'data-lake' in the background."

{
  "name": "run_background_task",
  "arguments": {
    "vm_name": "data-lake",
    "command": "python ingest.py"
  }
}

9. Deploy Configuration

"Upload my local .env.production file to /app/.env on the 'worker-node' VM."

{
  "name": "upload_file",
  "arguments": {
    "vm_name": "worker-node",
    "source": ".env.production",
    "destination": "/app/.env"
  }
}

10. Locate Error Logs

"Search for any files named error.log inside the /var/log directory."

{
  "name": "search_files",
  "arguments": {
    "vm_name": "monitor",
    "query": "error.log",
    "path": "/var/log"
  }
}

11. Setup File Watcher

"Configure a file sync. Map my local ./src folder to /usr/src/app on 'dev-main'."

{
  "name": "configure_sync",
  "arguments": {
    "vm_name": "dev-main",
    "host_path": "./src",
    "guest_path": "/usr/src/app",
    "direction": "bidirectional"
  }
}

12. Conflict Resolution

"There's a sync conflict on README.md. Use my local version."

{
  "name": "resolve_conflict",
  "arguments": {
    "vm_name": "docs-site",
    "file_path": "README.md",
    "resolution": "use_host"
  }
}

13. Live Debugging

"Show me the last 50 lines of the nginx error log on the 'proxy' VM."

{
  "name": "tail_vm_log",
  "arguments": {
    "vm_name": "proxy",
    "path": "/var/log/nginx/error.log",
    "lines": 50
  }
}

14. Monitor Background Task

"What is the output so far for task task_12345?"

{
  "name": "get_task_output",
  "arguments": {
    "vm_name": "data-lake",
    "task_id": "task_12345"
  }
}

15. Search Logs

"Search the syslog on 'auth-service' for any 'segfault' errors."

{
  "name": "grep_log_stream",
  "arguments": {
    "vm_name": "auth-service",
    "path": "/var/log/syslog",
    "pattern": "segfault"
  }
}

16. Pre-Update Backup

"I'm about to upgrade the database. Save a snapshot called 'before-v14-upgrade'."

{
  "name": "snapshot_save",
  "arguments": {
    "vm_name": "postgres-primary",
    "snapshot_name": "before-v14-upgrade"
  }
}

17. Disaster Recovery

"The upgrade failed! Restore to the 'before-v14-upgrade' snapshot."

{
  "name": "snapshot_restore",
  "arguments": {
    "vm_name": "postgres-primary",
    "snapshot_name": "before-v14-upgrade"
  }
}

18. List Snapshots

"What snapshots are available for the 'kafka-broker' VM?"

{
  "name": "snapshot_list",
  "arguments": {
    "vm_name": "kafka-broker"
  }
}

19. Investigate High Load

"The 'ml-trainer' VM is slow. List the running processes."

{
  "name": "list_processes",
  "arguments": {
    "vm_name": "ml-trainer"
  }
}

20. Kill Stuck Process

"Process ID 9982 is stuck on 'worker-01'. Kill it."

{
  "name": "kill_process",
  "arguments": {
    "vm_name": "worker-01",
    "pid": 9982,
    "signal": "SIGKILL"
  }
}

21. Check Service Availability

"Is port 8080 open and listening on the 'jenkins' VM?"

{
  "name": "check_vm_port",
  "arguments": {
    "vm_name": "jenkins",
    "guest_port": 8080
  }
}

22. System Health Dashboard

"Get me a full dashboard with CPU, RAM, and disk usage."

{
  "name": "get_vm_dashboard",
  "arguments": {
    "vm_name": "production-replica"
  }
}

23. Initiate Large Download

"Download the 10GB dataset from example.com to /data/ on the 'ai-model' VM."

{
  "name": "start_download",
  "arguments": {
    "vm_name": "ai-model",
    "url": "http://example.com/data.tar.gz",
    "destination": "/data/data.tar.gz"
  }
}

24. Blocking Wait

"Wait for the download operation op_5592 to finish."

{
  "name": "wait_for_operation",
  "arguments": {
    "operation_id": "op_5592",
    "timeout_seconds": 600
  }
}

25. Detect Zombie VMs

"Scan the system to see if we have any orphaned Zombie VMs."

{
  "name": "scan_system_health",
  "arguments": {}
}

26. Clean Zombies (Dry Run)

"Check what would happen if we cleaned up 'zombie-1' and 'old-test'."

{
  "name": "cleanup_zombies",
  "arguments": {
    "vm_names": ["zombie-1", "old-test"],
    "dry_run": true
  }
}

⚙️ MCP Configuration

Claude Desktop / Cline / Cursor

Add to your claude_desktop_config.json or mcp_config.json:

{
  "mcpServers": {
    "vagrant-mcp": {
      "command": "node",
      "args": ["C:\\path\\to\\Virtualbox-mcp-server\\apps\\mcp-server\\dist\\index.js"],
      "env": {
        "LOG_LEVEL": "info",
        "PATH": "C:\\Program Files (x86)\\Vagrant\\bin;C:\\Program Files\\Oracle\\VirtualBox;%PATH%"
      }
    }
  }
}

🔥 Top 5 MCP Configuration Examples

1. Development Environment (Node.js + Docker)

{
  "mcpServers": {
    "vagrant-mcp": {
      "command": "node",
      "args": ["/home/user/Virtualbox-mcp-server/apps/mcp-server/dist/index.js"],
      "env": {
        "LOG_LEVEL": "debug",
        "VAGRANT_HOME": "/home/user/.vagrant.d",
        "VM_DEFAULT_BOX": "ubuntu/jammy64",
        "VM_DEFAULT_MEMORY": "4096",
        "VM_DEFAULT_CPU": "4"
      }
    }
  }
}

2. CI/CD Pipeline (Jenkins/GitHub Actions)

{
  "mcpServers": {
    "vagrant-mcp": {
      "command": "node",
      "args": ["/opt/mcp/vagrant-mcp-server/dist/index.js"],
      "env": {
        "LOG_LEVEL": "warn",
        "VAGRANT_HOME": "/var/lib/jenkins/.vagrant.d",
        "VM_AUTO_DESTROY": "true",
        "VM_SNAPSHOT_BEFORE_TEST": "true"
      }
    }
  }
}

3. Windows Workstation

{
  "mcpServers": {
    "vagrant-mcp": {
      "command": "node.exe",
      "args": ["C:\\Users\\Developer\\mcp\\Virtualbox-mcp-server\\apps\\mcp-server\\dist\\index.js"],
      "env": {
        "LOG_LEVEL": "info",
        "PATH": "C:\\Program Files (x86)\\Vagrant\\bin;C:\\Program Files\\Oracle\\VirtualBox;C:\\Windows\\System32",
        "VAGRANT_HOME": "C:\\Users\\Developer\\.vagrant.d"
      }
    }
  }
}

4. macOS with Homebrew

{
  "mcpServers": {
    "vagrant-mcp": {
      "command": "/opt/homebrew/bin/node",
      "args": ["/Users/dev/Projects/Virtualbox-mcp-server/apps/mcp-server/dist/index.js"],
      "env": {
        "LOG_LEVEL": "info",
        "PATH": "/opt/homebrew/bin:/usr/local/bin:/usr/bin",
        "VAGRANT_HOME": "/Users/dev/.vagrant.d"
      }
    }
  }
}

5. Production/Enterprise (Restricted Environment)

{
  "mcpServers": {
    "vagrant-mcp": {
      "command": "node",
      "args": ["/srv/mcp/vagrant-server/dist/index.js"],
      "env": {
        "LOG_LEVEL": "error",
        "VAGRANT_HOME": "/srv/vagrant",
        "VM_MAX_COUNT": "10",
        "VM_ALLOWED_BOXES": "company/base-ubuntu,company/base-centos",
        "VM_REQUIRE_SNAPSHOT": "true",
        "GUARDRAILS_STRICT": "true"
      }
    }
  }
}

🧪 Development

# Watch mode (rebuild on changes)
npm run dev

# Lint
npm run lint

# Format
npm run format

📝 License

MIT © usemanusai

Made with ❤️ for AI-powered infrastructure management

⬆ Back to Top