Skip to content
Galois Docs

Quickstart

Get galois-edge running on your machine and controlling instruments in under 5 minutes. The daemon is a single self-contained binary — no Python runtime, no Docker, no dependencies.

Prerequisites

Section titled “Prerequisites”
  • Linux (x86_64 or arm64) or Windows 10/11
  • At least one instrument connected via USB, GPIB, LAN, or Serial

Step 1 — Install

Section titled “Step 1 — Install”
Terminal window
curl -fsSL https://get.galoislabs.ai | sh

The installer places galois-edge in /usr/local/bin and registers a systemd service. Supports x86_64 and arm64 (including Raspberry Pi).

Step 2 — Start the daemon

Section titled “Step 2 — Start the daemon”
Terminal window
galois-edge start

The daemon auto-discovers connected instruments and prints what it found:

galois-edge v0.9.4 starting...
  Scanning GPIB bus...    found Keithley 2400 at GPIB::24
  Scanning USB...         found Keysight 34461A at USB0::...
  gRPC API listening on   :50051
  WebSocket API on        :8765
  MCP server on           :8767/mcp
  Ready.

Step 3 — Run your first measurement

Section titled “Step 3 — Run your first measurement”

Point any MCP-capable agent at the daemon and every connected instrument shows up as a typed tool. Claude Desktop, Cursor, and any other client that reads mcpServers config takes the same shape:

// claude_desktop_config.json (Claude Desktop)
// ~/.cursor/mcp.json           (Cursor)
{
  "mcpServers": {
    "galois-edge": {
      "url": "http://localhost:8767/mcp"
    }
  }
}

Restart the agent. It now sees per-instrument typed tools (keithley_2400__source_voltage(value), keysight_34461a__measure_voltage_dc()) plus generic list_instruments, get_capabilities, start_sweep, and start_stream tools. See Agents (MCP) for the full client matrix.

Next steps

Section titled “Next steps”