VMware MCP Server

MCP server for managing VMware Workstation virtual machines. It exposes VM lifecycle, power, health, and snapshot operations as MCP tools that can be consumed by any MCP-compatible AI assistant.

It talks to VMware Workstation through two backends:

• vmrest.exe (HTTP REST API) — used for listing VMs, power control, configuration, network info, and health checks. Can be local or remote.

• vmcli.exe (local binary) — used for snapshot queries, which the REST API does not expose. Local-only: snapshot tools require the server to run on the same host as VMware Workstation.

Available Tools

Related MCP server: proxmox-mcp

Prerequisites

• Python >= 3.11

• VMware Workstation Pro 17+ with the REST API service enabled (vmrest.exe running on port 8697 by default)

• vmcli.exe (bundled with VMware Workstation) — required only for the snapshot tool, and only when the server runs on the same host as Workstation

• uv (recommended) or pip

Quick Start

# Clone / open the project directory
cd VMwareMCP

# Create a virtual environment and install
uv venv
uv pip install -e .

# Configure the vmrest host (use your Windows credentials)
export VMREST_HOST=localhost
export VMREST_USERNAME=your_username
export VMREST_PASSWORD=your_password

# Run the server
uv run vmware-mcp

Environment Variables

Example

# Using a .env file (copy .env.example to .env)
VMREST_HOST=localhost
VMREST_PORT=8697
VMREST_USERNAME=Admin
VMREST_PASSWORD=secret
VMCLI_PATH=C:\Program Files (x86)\VMware\VMware Workstation\vmcli.exe

Transport

The server defaults to streamable-http on port 51001.

• Server URL: http://localhost:51001/mcp

• Override with --transport stdio for stdio mode

MCP Client Configuration

streamable-http (default)

{
  "mcpServers": {
    "vmware": {
      "url": "http://localhost:51001/mcp"
    }
  }
}

stdio

{
  "mcpServers": {
    "vmware": {
      "command": "uv",
      "args": ["run", "vmware-mcp", "--transport", "stdio"]
    }
  }
}

Verifying the Setup

1. Ensure vmrest.exe is running: browse to http://localhost:8697/api/vms.

2. Start the server: uv run vmware-mcp.

3. Connect your MCP client to http://localhost:51001/mcp.

4. Call check_server_health first — it confirms connectivity and auth before you run other operations.

Notes on IDs vs Paths

• list_vms returns both an opaque id and the .vmx filesystem path.

• Pass the id to all tools (including get_vm_snapshots). The snapshot tool automatically resolves the id to the .vmx path via the vmrest API before invoking vmcli.exe — you never need to supply a path manually.

Project Structure

VMwareMCP/
├── src/vmware_mcp/
│   ├── server.py            # FastMCP server entry point + client accessors
│   ├── vmrest_client.py     # HTTP client for the vmrest REST API
│   ├── vmcli_client.py      # Subprocess client for the local vmcli.exe binary
│   ├── models.py            # Pydantic data models
│   ├── config.py            # Environment-based configuration
│   └── modules/
│       ├── vm.py            # VM list and detail tools
│       ├── power.py         # Power state query and control tools
│       ├── health.py        # Server/backend health check tool
│       └── snapshot.py      # Snapshot listing tool (via vmcli)
├── project-knowledge/       # Detailed architecture & contributor docs
├── pyproject.toml
└── README.md

For deeper architecture, setup, and contribution details, see the project-knowledge/ directory.

License

MIT