Clickclickclick

ClickClickClick is an open source framework from InstaVM for autonomous Android and computer use with any large language model, local or remote. Given a plain language task, such as drafting an email, finding places on a map in the browser or starting an online chess game, it drives the device's interface step by step by pairing a planner model that decides what to do with a finder model that locates the right UI element on screen. It is aimed at developers and researchers experimenting with agents that operate phones and desktops. It runs on Android by default and on macOS, and supports OpenAI and Gemini models as well as local models through Ollama; the maintainers report the best results with a Gemini Flash Lite model serving as both planner and finder. Tasks can be run from a command line tool, a Gradio web interface, a Python API or a REST API, with model settings kept in a YAML file and an image quality option that trades screenshot detail for speed. The framework is free under the MIT licence; users pay only for any hosted model API they connect. The listed website, instavm.io, belongs to InstaVM's separate commercial sandbox service, which has its own published plans: Free at $0 base plus pay as you go usage with $50 in starting credits, Pro at $100 per month base plus usage, and custom Enterprise pricing. The maintainers describe the code as highly experimental and advise using it at your own risk; local Ollama models are noted as unreliable for element detection.