Personal project
Arlene — Cross-Platform AI Assistant
I am building Arlene, a robust and deeply customizable AI assistant for the desktop, with a drag-and-drop agent pipeline designer that works with local and cloud providers.
Open-source release planned
The framework is not open source yet. I plan to publish its source; no release date is announced.
What I worked on
Arlene is the desktop AI assistant I always wanted to have: one cross-platform application, built with Electron, that treats the agent — not a single chat box — as the unit of work. I am developing it to be robust enough for daily use and plan to open-source it.
Its centre is a visual pipeline designer in the spirit of n8n: agents, tools, and steps are arranged with drag and drop into flows you can actually read and reuse. Arlene talks to local models and cloud providers through the same interface, so a pipeline is not married to one vendor.
It ships with built-in harnesses for many different tasks, and almost everything can be customized in depth — providers, tools, prompts, and behaviour. Self-learning and self-improvement features let it refine how it works over time instead of starting from zero in every session.
What it brings together
- Visual agent pipeline designer (drag & drop)
- Local and cloud model providers
- Built-in harnesses for many task types
- Deep customization of tools and behaviour
- Self-learning & self-improvement
- Windows, Linux & macOS from one codebase