Skip to main content
Glama
Amigent – Amiga remote control for AI agents

Amigent – Amiga remote control for AI agents

     _____            .__                            __
    /  _  \    _____  |__|   ____    ____    ____  _/  |_
   /  /_\  \  /     \ |  |  / ___\ _/ __ \  /    \ \   __\
  /    |    \|  Y Y  \|  | / /_/  >\  ___/ |   |  \ |  |
  \____|__  /|__|_|  /|__| \___  /  \___  >|___|  / |__|
          \/       \/     /_____/       \/      \/

     complete remote control for a real Amiga, by AI or manually
            tuned for slow Amigas and slow networks

Amigent connects your PC or Mac to your real Amiga over the home network. Then you can:

  • let an AI agent do the work, for example Claude Code or Codex: install software, update programs, change settings, copy files, click through programs. Just tell it what you want.

  • use the Amiga yourself from your computer, as a remote desktop like TeamViewer or VNC: a window with the live screen, your own mouse and keyboard, file transfer and a shared clipboard.

  • and both of these quickly, even on a slow machine. Amigent is tuned for slow Amigas and slow networks: a 7 MHz 68000 sends its whole screen in under a second, and over a plipbox a click shows its result in a fraction of a second. How it stays fast.

Both at the same time: you see every click the agent makes and can take over at any moment.

Amigent with a Workbench screen and the keyboard panel

Installation

On the Amiga

1. Copy amigentd to the Amiga (amigentd.000 on a 68000), or unpack the archive from the releases.

2. Start it with amigentd SETUP in a Shell, or double-click its icon.

On the PC or Mac

3. Paste this into your AI agent:

Install Amigent for me: run "pipx install amigent" (if pipx or Python 3.10
or newer is missing, install them first: Windows winget, macOS Homebrew,
Linux the package manager). Then run amigent-setup; it finds my Amiga on
the network by itself. Register the MCP server "amigent" with the command
amigent-server, and run amigent-viewer --install so the remote desktop
shows up as a program. Tell me if you need to be restarted.

The agent sets up the Amiga over the network; from then on amigentd starts with it.

Windows: Amigent appears in the Start menu. For an encrypted connection openssl is needed, which comes with Git for Windows.

macOS: Amigent appears in ~/Applications. On its first start, allow it access to the local network, or it cannot reach the Amiga. Cmd is the right Amiga key.

Linux: there is no program entry; start the remote desktop with amigent-viewer.

Without an AI agent: see the manual installation.

Related MCP server: agentic-remote-pc

What you need

  • An Amiga with AmigaOS 2.04 or newer, any CPU from the 68000 on, a network connection (a plipbox on the parallel port is enough) and a TCP/IP stack such as Roadshow.

  • A PC or Mac with Windows, macOS or Linux and Python 3.10 or newer.

What you can ask for

"Check which programs on the Amiga are outdated."
"Install the game from my Downloads folder and add it to the Games menu."
"Copy the demo to the Amiga and start it."
"Open the remote desktop."

Made for slow machines and slow wires

Most of the work in Amigent went into making a slow Amiga on a slow network feel quick, for the AI agent and for you at the mouse:

  • The screen travels as the Amiga keeps it. Chipset screens go out as bitplanes, graphics-card screens as small palette tiles, and only the tiles that changed, often as the difference to what you already have.

  • Your input comes first. Mouse moves, clicks and keys are carried out even while a picture is still on its way, and the area around the pointer is sent before the rest.

  • The Amiga keeps its time. While nothing happens, Amigent asks rarely and costs the Amiga only a few percent of its CPU.

  • Files go in one go. Whole folders travel in a single stream, and files that are already there stay where they are.

Measured with a 640×256 Workbench screen:

Amiga

whole screen

unchanged screen

CPU while watching

68000, 7 MHz

0.6–1.0 s

80–95 ms

under 10 %

A1200, 68020 at 14 MHz

0.1 s

15 ms

2–3 %

On a 1920×1080 graphics-card screen over a plipbox (about 50 KB/s):

Seconds per screen update over a plipbox, bar lengths to scale: whole screen as raw pixels 124 s each time; Amigent 1.9 s for the first picture (64 times faster), 170 ms for a typed line (730 times), 7 ms when nothing changed (18,000 times)

The orange bars are what moving the whole screen as raw 24-bit pixels costs on that wire, every time; the blue ones are Amigent, measured on a real A1200. All numbers, how they were taken and how to repeat them are in docs/PERFORMANCE.md.

Good to know

  • New address after a restart? Amigent finds the Amiga again by itself. It takes only the one that proves to be yours.

  • Several Amigas? amigent-fleet shows them side by side with a preview.

  • After a big OS update, amigent-repair puts back what the update removed.

  • Safe at home. A key keeps other computers on the network out, and the connection can be encrypted. Don't open the Amiga to the internet.

More in the guide and the FAQ.

License

MIT, see LICENSE. By Andreas Wendorf (Akustikrausch). Changes are in the changelog.

Related MCP Connectors

Related MCP Servers

  • A
    license
    A
    quality
    C
    maintenance
    Enables AI agents to interact with Classic Mac OS 7-9 through structured commands and responses using a shared folder transport, supporting file browsing, window management, mouse/keyboard control, and more.
    23
    41 npm
    28
    GPL 3.0
  • A
    license
    Not graded
    quality
    C
    maintenance
    Enables AI agents to see and control an existing Linux X11 desktop over SSH by taking screenshots and performing mouse, scroll, typing, and keyboard-shortcut actions across browsers, terminals, file managers, and other apps.
    MIT
  • A
    license
    Not graded
    quality
    C
    maintenance
    Enables AI clients to securely control trusted devices and perform filesystem, terminal, process, and document operations through a self-hosted remote service.
    724,305 npm
    MIT