protoAgent beta

Download

protoAgent on your desktop

v0.191.0 beta 2026-10-03
Download for macOS (.dmg) Windows (setup.exe)

98 MB · Apple Silicon · signed & notarized · updates itself in place.

Download for Windows (setup.exe) macOS (.dmg)

98 MB · Windows 11 x64 · unsigned build · updates itself in place.

Download for Linux (.AppImage) Debian/Ubuntu (.deb)

206 MB AppImage (128 MB .deb) · x86_64 · glibc 2.35+ · unsigned build · the AppImage updates itself in place.

macOS (.dmg) Windows (setup.exe) Linux (.AppImage)

Couldn’t detect your platform — pick a build, or run it from source.

System requirements — macOS

Apple Silicon (M1 or later)
macOS 13 Ventura or newer
A model gateway — connect an OpenAI-compatible endpoint or key in setup

Installation

1
Download the .dmg above.
2
Open it and drag protoAgent to Applications.
3
Launch protoAgent — it starts its own bundled server. No Python, Node, or runtime is fetched at first launch; the installed app is about 100 MB, all in.
4
The setup wizard connects your model and you’re live.
Signed & notarized by Apple
Opens cleanly, no Gatekeeper workaround. protoAgent is in public beta and updates itself in place; expect frequent releases.

System requirements — Windows

Windows 11, 64-bit (x64)
A model gateway or subscription — connect an OpenAI-compatible endpoint, an API key, or your ChatGPT/Claude plan in setup

Installation

1
Download the setup.exe above and run it.
2
Launch protoAgent — it starts its own bundled server, nothing else to install.
3
The setup wizard connects your model and you’re live.
Unsigned build — expect a SmartScreen prompt
There’s no Windows code-signing certificate yet, so both the installer and the first launch trigger "Windows protected your PC." Click More info → Run anyway — that’s expected for this build, not a sign of tampering. protoAgent is in public beta and updates itself in place; expect frequent releases.
Full install, data-location & recovery guide →

System requirements — Linux

x86_64, glibc 2.35 or newer (Ubuntu 22.04+, Fedora 36+, Debian 12+)
A desktop session — GNOME, KDE, or any compositor, on Wayland or X11
A model gateway or subscription — connect an OpenAI-compatible endpoint, an API key, or your ChatGPT/Claude plan in setup

Installation

1
Download the .AppImage above, chmod +x it, and run it — or install the .deb with apt.
2
Launch protoAgent — it starts its own bundled server, nothing else to install.
3
The setup wizard connects your model and you’re live.
Needs a real desktop session
The app renders through WebKitGTK, which requires working accelerated compositing. In a normal GNOME/KDE session on Wayland or X11 that’s a given. Over a display without GPU acceleration — a bare Xvfb, some container and CI setups, X-forwarding or VNC without DRI3 — WebKitGTK 2.52 crashes on launch (#2866). Run the server headless there instead: python -m server --ui console. protoAgent is in public beta and the AppImage updates itself in place; a .deb install is apt’s job.

Get release announcements

The desktop app ships for macOS, Windows, and Linux, and updates itself in place. Leave your email if you’d rather hear about releases than watch the changelog.

No spam — release announcements only. Unsubscribe anytime.

Prefer to run it yourself?

protoAgent is open source — clone the repo and run the console + server locally (npm ci then python -m server). The first-agent tutorial walks through it. Runs anywhere Python does — macOS, Linux, or Windows.

View full release history →