ActingCommand - Android emulator automation for AI agents
by•
You can start by reading the README. This project is still evolving rapidly. This release includes only the Runtime; the UI and its repository are not yet public. However, the Runtime already provides most of the project's functionality.
Replies
Best
Maker
📌
The idea is to separate planning from execution: an external AI agent or developer prepares tasks, while the Rust runtime handles scheduling, screen recognition, device actions, and execution records.
This launch focuses on the open-source Runtime. The UI is not yet public, and the project is still evolving rapidly. Please start with the README for setup instructions, current capabilities, and limitations.
I'd love feedback from people building AI agents or emulator automation tools. What repetitive workflow would you like to automate, and what would make integration easier for you?
What became possible in your product with Astra that was not practical before?
Maker
I've used Codex since GPT-5.5. I'm not a programmer; I mostly describe what I want and work with the model to make it happen.
GPT-5.5 helped me discuss and implement my ideas. GPT-5.6 Sol built a useful foundation, but felt too cautious and slow. In my experience, it spent too many tokens on tests that checked past mistakes rather than anticipating new problems, making progress costly.
Astra restored momentum. The biggest improvement for my project has been visual recognition and spatial understanding in games and apps, especially in 3D environments where agents need stronger judgment.
ActingCommand's goal is to let an agent play or run a workflow once, capture what happened in a ledger and logs, then turn that experience into reusable automation and refine it over time. Astra makes that approach more practical for me: it handles situations Sol struggled with and helps me iterate faster. As a non-programmer, I can now move this project forward much more effectively.
Replies