# Symbioza

Runs containerized GPU jobs under a strict budget ceiling and retrieves generated files. Useful for fine-tuning models and batch inference without surprise bills.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI
- Author: symbioza
- License: MIT · policy=link-only · description only; the file stays with the author
- Source: https://github.com/symbioza/claude-plugin
- Updated: 2026-10-11
- HTML: https://affpapa.org/en/nevibecode/mcp/symbioza

## What it does

Symbioza links your coding assistant to cloud GPU nodes through a remote HTTP endpoint. The agent specifies a container image and an execution command. A strict spending ceiling applies to every run. Hardware provisioning and artifact retrieval happen automatically.

The server provides nine tools, including `estimateExecution`, `submitJob`, and `getArtifact`. You can inspect projected runtimes and hourly rates before spending money. Estimation is completely free.

Authentication relies on browser OAuth. No static API token is required. Submissions will fail if your account balance is zero, so you must add funds before dispatching compute.

## Requirements
- Needs: Docker
- Remote server: `https://symbioza.dev/mcp`

## Install

### Claude Code (native platform)

**Platform CLI: remote server**
```bash
claude mcp add --transport http symbioza https://symbioza.dev/mcp
```

### Codex

**Platform CLI: remote server**
```bash
codex mcp add symbioza --url https://symbioza.dev/mcp
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "symbioza": {
            "url": "https://symbioza.dev/mcp"
        }
    }
}
```

### GitHub Copilot

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

### Gemini CLI

**Platform CLI: remote server**
```bash
gemini mcp add -t http symbioza https://symbioza.dev/mcp
```

## When to use it
- Cap cloud GPU spending with a hard financial cutoff
- Preview training runtime and expenses prior to execution
- Launch containerized model evaluations directly from coding agents
- Download generated checkpoints and logs after a remote run completes

## Example prompts
- Estimate the cost to run an evaluation job in `pytorch/pytorch:latest` before starting compute.
- Check the progress of my Symbioza job and download the saved weights once finished.
- Draft the container arguments for batch inference on Symbioza without submitting yet.

## Similar
- [MCP Toolbox for Databases](https://affpapa.org/en/nevibecode/mcp/googleapis-mcp-toolbox-for-databases.md) — MCP server · ★ 16741
- [labelhead-artist-momentum](https://affpapa.org/en/nevibecode/mcp/labelhead-artist-momentum.md) — MCP server · ★ 108044
- [Agent Skills Search Server](https://affpapa.org/en/nevibecode/mcp/agent-skills-search-server.md) — MCP server · ★ 26711
- [Skyvern](https://affpapa.org/en/nevibecode/mcp/skyvern.md) — MCP server · ★ 23288
- [Nhost](https://affpapa.org/en/nevibecode/mcp/nhost.md) — MCP server · ★ 9370
- [PraisonAI](https://affpapa.org/en/nevibecode/mcp/praisonai.md) — MCP server · ★ 9301

---
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