# mnvitop

One terminal for the GPUs of all your ssh hosts. It runs `nvidia-smi` over ssh
and shows every host's GPUs, their memory, utilisation history and running
processes in a single live view. Nothing needs to be installed on the remote
side beyond `nvidia-smi` and `ps`. It also ships an MCP server, so an agent can
ask where a free GPU is.

![mnvitop watching three GPU hosts](https://raw.githubusercontent.com/kangkangzi2025/mnvitop/main/docs/demo.gif)

## Install

```bash
uv tool install mnvitop
```

You need Python 3.11 or newer locally and ssh access to the hosts. To get the
MCP server as well, install `mnvitop[mcp]`.

## Use

```bash
mnvitop h20 semi pro6000
```

Each argument is an ssh destination, either an alias from `~/.ssh/config` or
`user@host`. Ctrl-C quits.

```bash
mnvitop -a        # temperature, process cpu and memory, more processes per GPU
mnvitop --once    # print once and exit
mnvitop --json    # one JSON line per round, for logging
```

`mnvitop --help` lists the rest. To stop typing the hosts every time, put them
in `~/.config/mnvitop/hosts.toml`, following
[hosts.example.toml](https://github.com/kangkangzi2025/mnvitop/blob/main/hosts.example.toml).

For agents, `mnvitop-mcp` is a stdio MCP server with three tools: `find_gpus`,
`fleet_status` and `list_hosts`. Register it in Claude Code with
`claude mcp add mnvitop -- mnvitop-mcp`. The repository also carries a matching
skill in `skills/gpu-fleet`.

MIT license.