@scratchrun/mcp-server

MCP server for ScratchRun — ephemeral, MicroVM-isolated code execution for AI agents.

Each call runs in a fresh hardware-isolated MicroVM. The VM is hard-purged after execution. No state, no files, nothing persists between calls.

Install

Add to your claude_desktop_config.json:

config.json

{
  "mcpServers": {
    "scratchrun": {
      "command": "npx",
      "args": ["-y", "@scratchrun/mcp-server"],
      "env": {
        "SCRATCHRUN_API_KEY": "sr_live_your_key_here"
      }
    }
  }
}

Get an API key at scratchrun.dev.

Tool: `scratchrun_exec`

Executes code in an ephemeral sandbox and returns stdout, stderr, exit code, and any output files.

Parameters:

Parameter	Type	Required	Description
`runtime`	`python3.12` \| `python3.11` \| `node20` \| `bash`	yes	Runtime to use
`code`	string	yes	Code to execute
`timeout_ms`	integer	no	Timeout in ms (default 10000, max 30000)
`memory_mb`	integer	no	Memory limit in MB (default 256, max 512)
`env`	object	no	Environment variables — use for secrets, not code strings
`files`	object	no	Files to write before execution (path → content)
`return_files`	string[]	no	File paths to capture after execution (returned as base64)

Example — run Python and return a chart:

config.json

{
  "runtime": "python3.12",
  "code": "import matplotlib.pyplot as plt\nimport numpy as np\nx = np.linspace(0, 10, 100)\nplt.plot(x, np.sin(x))\nplt.savefig('/tmp/plot.png')",
  "return_files": ["/tmp/plot.png"]
}

Output files are returned as base64. Image files (PNG, JPG, SVG) are returned as MCP image content blocks and render inline in Claude.

Isolation

Hardware-virtualized MicroVM per execution (own kernel, not a shared-kernel container)
TerminateMicroVM called unconditionally after every run — VM destroyed, never reused
RFC 1918 + cloud metadata (169.254.x.x) always blocked at the network layer
Read-only system filesystem via OverlayFS; /tmp is RAM-backed and gone with the VM

Latency

~400ms median (warm pool). 2–8s cold start if the pool is empty.

License

MIT