Uindow

Uindow is an automation platform designed to replicate human browser interactions at the operating system level, enabling users to automate web workflows in a way that is indistinguishable from real human activity. It addresses the challenge of automating tasks on websites that employ bot detection by generating genuine mouse movements, keystrokes, and scrolls, all executed from the user's own computer and IP address. This approach ensures that automations are not flagged as synthetic or blocked by anti-bot measures.

The tool features a built-in integrated development environment (IDE) for modeling workflows as finite-state machines, supporting complex branching, retries, and loops to handle dynamic web behavior. Users can record their actions directly, with Uindow translating these into deterministic JavaScript, eliminating the need for manual scripting or syntax learning. Automations can be exported and shared as portable YAML files, with module inputs kept separate from source code to protect sensitive data. The platform also provides a module repository for discovering, installing, and sharing automation modules within the community, and allows for private sharing outside its repository.

Uindow offers a range of capabilities for both technical and non-technical users. It supports precise element selection with a custom CSS selector engine, open-sourced for transparency. The platform includes features for scheduling automations, persistent and session-based data storage, and human-in-the-loop steps for tasks requiring manual intervention. Every automation run generates comprehensive evidence, including screenshots, video, logs, and data exports, all saved locally for auditing and debugging. Security and privacy are emphasized: automations run in isolated profiles, local services are locked down, and user credentials never leave the machine. No usage data is tracked or collected by the platform itself.

Integration with AI agents such as Claude, Cursor, Copilot, or any MCP-compatible agent is supported, enabling local, offline control. Uindow provides agents with both screenshot-based vision and a viewport-aware list of on-screen elements for accurate interaction, even on visually complex or obfuscated sites. The tool can be operated from the command line or integrated into CI/CD pipelines via @uindow/cli, with support for live log streaming and a built-in local proxy for data access. All installed modules come with fully auditable source code, ensuring transparency and user control over every aspect of automation.