# Nodegrove VRAM MCP server

> Determines if open-weight LLMs fit on specific GPUs and provides hardware recommendations and VRAM estimates.

- Repository: https://github.com/nodegrove/vram-mcp
- Page: https://hostedmcp.org/servers/nodegrove/vram-mcp
- Transport: stdio, remote
- Stars: 0
- License: MIT
- Latest version: 1.1.5

## Install

### Claude Code

```sh
claude mcp add vram-mcp -- npx -y @nodegrove/vram-mcp
```

### Cursor (`~/.cursor/mcp.json`)

```json
{
  "mcpServers": {
    "vram-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@nodegrove/vram-mcp"
      ]
    }
  }
}
```

### VS Code (`.vscode/mcp.json`)

```json
{
  "servers": {
    "vram-mcp": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@nodegrove/vram-mcp"
      ]
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"vram-mcp","type":"stdio","command":"npx","args":["-y","@nodegrove/vram-mcp"]}'
```

### Claude Desktop (`claude_desktop_config.json`)

```json
{
  "mcpServers": {
    "vram-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@nodegrove/vram-mcp"
      ]
    }
  }
}
```

### Codex

```sh
codex mcp add vram-mcp -- npx -y @nodegrove/vram-mcp
```

### Codex config.toml (`~/.codex/config.toml`)

```toml
[mcp_servers.vram-mcp]
command = "npx"
args = ["-y", "@nodegrove/vram-mcp"]
```

### Gemini CLI

```sh
gemini mcp add -s user vram-mcp npx -- -y @nodegrove/vram-mcp
```

### Gemini settings.json (`~/.gemini/settings.json`)

```json
{
  "mcpServers": {
    "vram-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@nodegrove/vram-mcp"
      ]
    }
  }
}
```

## Tools (4)

- `can_i_run`
- `what_fits`
- `estimate_vram`
- `estimate_from_hf_repo`

## FAQ

### How does the server calculate VRAM requirements?

It calculates memory as the sum of weights, KV cache, and overhead, using specific formulas for different model architectures like sliding-window or hybrid models.

### Is the server free to use?

Yes, the server is free, read-only, and does not require an account or an API key.
