Handbeam - An experimental open-source AI agent for mobile devices

Hi everyone! I’m building Handbeam, an experimental open-source AI agent for mobile devices, built with Elixir and an embedded Erlang/OTP runtime.

The goal is to explore what an agent designed for a phone can do. Coding is one use case, rather than the whole purpose of the project.

My motivation is that many mobile agent experiences seem to fall into two groups: proprietary assistants provided by phone manufacturers, or mobile interfaces to agents primarily designed for desktop and terminal workflows. I wanted to explore an open-source alternative built around the phone itself, rather than treating it mainly as a remote control for another machine.

Handbeam runs its agent loop and tools on the device, with a native mobile UI. LLM inference uses a configured API provider, so it is not a fully offline assistant. Current capabilities include working with local files, running Elixir scripts, and using Git. An embedded Elixir/Erlang environment also makes it possible to experiment with executable tools and small projects directly on the phone.

The project is still experimental. Android has been tested on a physical device; iOS has been tested in the simulator. Code executes in the app’s BEAM VM, not in a separate sandbox.

I’d love to hear what the Elixir community thinks about this direction. What would you want a phone-native agent to do beyond chat and coding?