The portability angle is compelling. The interesting design problem to me is whether a resumable session artifact has to be identical to the raw transcript.
I'd love a middle layer that preserves enough state to resume and debug while aggressively quarantining obvious secret classes and high-risk blobs. Right now the tradeoff is basically perfect replayability versus shareability, and it feels like there should be a useful middle ground there.
I've been frustrated with my AI coding workflow lately. Switching between different browser tabs and separate CLI tools for different models (Claude, Gemini, OpenAI) was constantly breaking my flow. I also wanted a way to interact with these tools away from my desk without the pain of using a standard mobile keyboard for CLI tasks. （Screenshots are available in the GitHub README linked above.）
So, I built VibeAround to scratch my own itch. It mainly solves two problems:
A multi-pane desktop workspace: It runs tools like Claude Code, Gemini, and OpenAI Codex side-by-side in a unified terminal. It behaves a bit like tmux for AI agents, letting you run tasks in parallel and compare outputs with all your context in one view.
A usable mobile terminal experience: I built a responsive mobile Web UI that connects to the local environment (e.g., via an ngrok tunnel). To make CLI work actually usable on a phone screen, I added a custom virtual keypad with essential keys like Ctrl+C, Esc, Tab, and a dedicated prompt button.
The goal is to turn your local machine into a private, multi-model AI server that you can control seamlessly from any screen, rather than relying on a single vendor's remote environment.
It is still very much a work in progress, but the core workflow is there and I'm using it daily. I'd love to hear your thoughts on the approach. How are you all managing multi-model CLI workflows locally? Happy to answer any questions about the implementation!