# VoltageGPU MCP server

> Deploy and manage confidential VMs and standard GPU pods with Intel TDX and NVIDIA attestation via a command-line interface or Python SDK.

- Repository: https://github.com/VoltageGPU/VOLTAGEGPU-CLI
- Page: https://hostedmcp.org/servers/voltagegpu/voltagegpu-cli
- Transport: —
- Stars: 1
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `VOLT_API_KEY` | no | no |  |

## FAQ

### How is VoltageGPU billed?

Billing is performed per second. For the first hour, you are charged upon deployment, but if you stop a standard pod, the unused portion of that hour is returned to your balance.

### How do I log in to the CLI?

You can create an API key on their website and then run 'voltagegpu login' to prompt for the key, or export the VOLT_API_KEY environment variable for scripts.

### Can I use the VoltageGPU MCP server with Claude?

Yes, you can add it to Claude Code using the command 'claude mcp add voltagegpu -e VOLT_API_KEY=volt_... -- vgpu mcp'.

### What is the difference between Confidential VMs and standard pods?

Confidential VMs use Intel TDX trust domains with NVIDIA GPUs and allow for self-generated attestation, while standard pods are whole GPU machines without an enclave for non-sensitive workloads.