OpenAdaptAn open-source framework automating desktop workflows using large multimodal models.

OpenAdapt is an open-source platform for GUI automation with machine learning. It records human demonstrations, trains models, and deploys agents that can operate any software, including automating tasks in various industries such as HR, law, insurance, healthcare, finance, logistics, and pharmacy. The platform uses multimodal models for automation, eliminating the need for prompt engineering, scripting, or brittle selectors. Built-in PII/PHI scrubbing keeps sensitive data safe. One command installs everything needed, and the platform works on Intel and Apple Silicon Macs. The user interface is simple, with commands like 'openadapt capture start' to start recording a task and 'openadapt doctor' to check the system. The platform has a fast iteration mode for development and debugging, and its ecosystem includes official benchmarks, a CLI, and a benchmark viewer for replaying and inspecting agent failures. Built on the uv tool, which is a fast Python package manager, OpenAdapt provides a simple and efficient way to create automation agents that can be used across various industries and software systems. OpenAdapt is a work in progress and is considered alpha software, meaning it comes with no warranty of any kind. To learn more about OpenAdapt and get started with it, users can visit the project's GitHub page and read the documentation on how to install and use the platform. OpenAdapt is a promising tool for automating tasks in various industries and software systems, and its open-source nature makes it accessible to developers and researchers who want to contribute to and customize the platform.