# Auxen

Spins up a private instance of an open model on a dedicated GPU from inside your agent chat. Billing is per minute.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI
- Author: auxen-ai
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/auxen-ai/auxen-mcp
- Updated: 2026-05-30
- HTML: https://affpapa.org/en/nevibecode/mcp/auxen

## What it does

Auxen is a remote MCP server at `https://api.auxen.ai/mcp` speaking StreamableHTTP. Point your client at that URL and the agent gets six tools. Four are read-only: `auxen_list_models` returns what is available, with an optional size filter; `auxen_list_instances` and `auxen_get_instance_status` report what is running and where; `auxen_get_balance` reads USD credits and active subscriptions.

The other two move real money. `auxen_provision_model` creates an instance, and `auxen_destroy_instance` tears one down for good. An agent that runs tool calls without asking will spend your balance on its own, so keep those two behind a confirmation step.

Auth comes in two flavours. Browser-based clients use OAuth 2.1 with PKCE and can self-register when they support Dynamic Client Registration. Headless agents send an `auxen_live_*` or `auxen_test_*` key as `Authorization: Bearer <key>` on every call; you generate it in the Auxen dashboard.

## Requirements
- Remote server: `https://api.auxen.ai/mcp`

## Install

### Claude Code (native platform)

**Platform CLI: remote server**
```bash
claude mcp add --transport http auxen https://api.auxen.ai/mcp
```

### Codex

**Platform CLI: remote server**
```bash
codex mcp add auxen --url https://api.auxen.ai/mcp
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "auxen": {
            "url": "https://api.auxen.ai/mcp"
        }
    }
}
```

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "auxen": {
            "type": "http",
            "url": "https://api.auxen.ai/mcp"
        }
    }
}
```

### Gemini CLI

**Platform CLI: remote server**
```bash
gemini mcp add -t http auxen https://api.auxen.ai/mcp
```

## When to use it
- run a private Llama or Qwen instance on a dedicated GPU
- check which models exist and what a minute costs
- fetch the endpoint and API key of a running instance
- read the credit balance before provisioning
- shut down an instance that is no longer needed

## Example prompts
- List the models available and their sizes
- How many credits are left on my account?
- Show the status and endpoint of the instance I started yesterday

## Similar
- [hyperloom-workload-optimizer](https://affpapa.org/en/nevibecode/skills/hyperloom-workload-optimizer.md) — skill · 785 installs
- [gke-inference](https://affpapa.org/en/nevibecode/skills/gke-inference.md) — skill · 6791 installs
- [llm-architect](https://affpapa.org/en/nevibecode/agents/llm-architect.md) — subagent · 649 installs
- [huawei-cloud-ascend-models-deploy](https://affpapa.org/en/nevibecode/skills/huawei-cloud-ascend-models-deploy.md) — skill · 424 installs
- [earth2studio-deterministic-forecast](https://affpapa.org/en/nevibecode/skills/earth2studio-deterministic-forecast.md) — skill · 1900 installs
- [llm-config](https://affpapa.org/en/nevibecode/skills/llm-config.md) — skill · 741 installs

---
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