Eaon Desktop

The Eaon desktop app — an AI chat client that runs on your own API keys. Built with Electron, React and TypeScript; the interface follows the supplied design frames.

Run it

npm install
npm run dev        # hot-reloading dev build
npm run build      # production bundle into out/
npm start          # run the production bundle
npm run dist:mac   # package a .dmg / .zip
npm run dist:win   # package the Windows installer (x64 + arm64)
npm run dist:linux # package an AppImage and a .deb (on Linux)

Install it

• macOS (12 or later): open Eaon-<version>.dmg and drag Eaon to Applications.

• Windows (10 or later, x64 or arm64): run Eaon-<version>-setup.exe. It installs for your user, with no administrator prompt. Windows SmartScreen may say "Windows protected your PC" until the installer is signed; choose More info, then Run anyway. Local models need Microsoft's Visual C++ Redistributable (arm64: this one), which most PCs already have.

• Linux (x64 or arm64): either

  • sudo apt install ./eaon-desktop_<version>_.deb, or
  • chmod +x Eaon-<version>-.AppImage && ./Eaon-<version>-.AppImage.

  Both run on Ubuntu 22.04 and later, Debian 12 and Fedora. The .deb also installs an AppArmor profile, which Ubuntu 23.10 and later need before Chromium's sandbox will start.

Bring your own key

Open Settings → Model providers and paste a key. Providers ship configured out of the box:

Provider	Notes
Anthropic	Official SDK, adaptive thinking, effort levels
OpenAI	Chat completions
Google Gemini	Via its OpenAI-compatible endpoint
OpenRouter, Groq	Chat completions
Ollama	Local, no key needed
Custom endpoint	Anything speaking the OpenAI chat-completions format

Keys are encrypted with the OS keychain (Electron safeStorage), written to your user-data directory, and never sent to the renderer process or anywhere except the provider you entered them for. After saving a key the app calls the provider's model list, so the picker fills in immediately.

Model lists come from a catalog generated from Pi's provider data and models.dev (npm run generate:models), checked against models.dev once a day while the app runs, and merged with each provider's own /models once a key is added. Removing a model in Settings → Model providers hides it, and it can be restored from the same page.

The composer's Effort control offers exactly the levels the chosen model takes, named the way providers name them: Off, Minimal, Low, Medium, High, Extra high and Max.

What's in the app

Three tabs, centred at the top of every screen (⌘1 / ⌘2 / ⌘3):