← Back to docs

MCP configuration

Connect your AI agent to Volspan GPUs using the Model Context Protocol.

What is MCP

The Model Context Protocol (MCP) is an open standard that lets AI agents call external tools. Volspan exposes GPU compute as MCP tools, so any MCP-compatible agent can submit code, check balances, and retrieve results without writing HTTP requests.

Supported clients

  • Claude Code
  • Claude Desktop
  • Cursor
  • Cline

Any client that supports the MCP server spec will work.

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json on macOS or the equivalent path on your OS.

{
  "mcpServers": {
    "volspan": {
      "command": "npx",
      "args": ["-y", "@volspan/mcp-server"],
      "env": {
        "VOLSPAN_API_KEY": "vc_live_your_key_here"
      }
    }
  }
}

Claude Code

Add to your project's .mcp.json at the repo root.

{
  "mcpServers": {
    "volspan": {
      "command": "npx",
      "args": ["-y", "@volspan/mcp-server"],
      "env": {
        "VOLSPAN_API_KEY": "vc_live_your_key_here"
      }
    }
  }
}

Cursor

Edit .cursor/mcp.json in your project root.

{
  "mcpServers": {
    "volspan": {
      "command": "npx",
      "args": ["-y", "@volspan/mcp-server"],
      "env": {
        "VOLSPAN_API_KEY": "vc_live_your_key_here"
      }
    }
  }
}

Cline

Open Cline settings and add a new MCP server with the same configuration as above.

Available MCP tools

run_on_gpu

Submit Python code to execute on a GPU. Returns run ID, output, and artifacts.

check_vault_balance

Check your current Vault balance in cents.

get_gpu_run

Get the status and output of a specific run by ID.

list_gpu_runs

List recent GPU runs with their status and cost.

Environment variables

VOLSPAN_API_KEY

Required. Your API key starting with vc_live_.

VOLSPAN_BASE_URL

Optional. Override the API endpoint. Defaults to https://api.volspan.dev.

Need help? Reach out at [email protected]