# Thunder Compute

Connects your coding agent to Thunder Compute cloud to spin up GPU machines and manage machine learning instances.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI · ★ 34
- Author: Thunder-Compute
- License: MIT · policy=link-only · description only; the file stays with the author
- Source: https://github.com/Thunder-Compute/thunder-cli
- Updated: 2026-09-16
- HTML: https://affpapa.org/en/nevibecode/mcp/thunder-compute

## What it does

This MCP server links your development agent with Thunder Compute infrastructure. Through this bridge, assistants can provision graphics instances, track expenses, handle snapshots, and inspect SSH keys.

The integration operates over an HTTP transport pointing to `https://api.thundercompute.com:8443/mcp`. The underlying Go-based `tnr` tool also manages port forwards, runs SCP transfers, and connects sessions.

Headless environments fail standard browser authentication callbacks on `127.0.0.1`, which requires creating an explicit API token in the web console first.

## Requirements
- Needs: a prebuilt binary from the repository releases
- Remote server: `https://api.thundercompute.com:8443/mcp`

## Install

### Claude Code (native platform)

**Platform CLI: remote server**
```bash
claude mcp add --transport http thunder-compute https://api.thundercompute.com:8443/mcp
```

### Codex

**Platform CLI: remote server**
```bash
codex mcp add thunder-compute --url https://api.thundercompute.com:8443/mcp
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "thunder-compute": {
            "url": "https://api.thundercompute.com:8443/mcp"
        }
    }
}
```

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "thunder-compute": {
            "type": "http",
            "url": "https://api.thundercompute.com:8443/mcp"
        }
    }
}
```

### Gemini CLI

**Platform CLI: remote server**
```bash
gemini mcp add -t http thunder-compute https://api.thundercompute.com:8443/mcp
```

## When to use it
- provision GPU instances for machine learning prototyping
- inspect compute budgets and billing details
- manage instance snapshots and configure SSH connections

## Example prompts
- Spin up a fresh GPU instance and report its status
- Check active snapshots and the current cloud billing balance
- Set up a remote compute instance for training experiments

## Similar
- [mcp-server](https://affpapa.org/en/nevibecode/mcp/mcp-server.md) — MCP server · ★ 43872
- [Coder](https://affpapa.org/en/nevibecode/mcp/coder.md) — MCP server · ★ 17160
- [firebase-mcp](https://affpapa.org/en/nevibecode/mcp/firebase-mcp.md) — MCP server · ★ 4481
- [Radar Kubernetes MCP Server](https://affpapa.org/en/nevibecode/mcp/radar-kubernetes-mcp-server.md) — MCP server · ★ 3905
- [Supabase](https://affpapa.org/en/nevibecode/mcp/supabase.md) — MCP server · ★ 2969
- [Microsoft Fabric MCP Server](https://affpapa.org/en/nevibecode/mcp/microsoft-fabric-mcp-server.md) — MCP server · ★ 3799

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md