Open-Interface

Why we included this project

Open Interface is for people who would rather describe a desktop task in plain language than write a script to do it. You tell it what you want, it sends the request to an LLM backend like GPT-4o or Gemini, breaks the job into concrete steps, and carries them out by simulating keyboard and mouse input. What sets it apart is the loop: it takes fresh screenshots as it works and sends them back to the model, so it can notice when something went wrong and correct course. That turns it from a one-shot automation into something closer to an autonomous assistant. It runs on macOS, Linux, and Windows, so developers and power users exploring self-driving desktops can try it on their own machines. It is early-stage and will need guardrails in real use, but for testing what agentic desktop control can do, it is a clear and approachable starting point.