OpenMuse is a personal agent application that combines browser automation, terminal access, file management, and persistent workspace capabilities built with CopilotKit.
The tool addresses the need for an AI agent that can execute multi-step tasks across multiple interfaces while maintaining visibility and control. It runs a persistent Chromium browser and optional Linux workspace on its own server, allowing the agent to browse public pages, execute terminal commands, manage files, and transfer documents between the agent's environment and the application. Users can ask for an outcome, review the agent's planned actions, and take manual control of the browser or terminal session when needed to continue work themselves.
OpenMuse is designed for self-hosting and customization rather than as a managed service. It works as a template that developers can clone and modify to their needs. The tool is currently in alpha and requires configuration for open-ended reasoning, live Google account integration, and CopilotKit Rich Threads. It runs on iOS, Android, and web platforms through CopilotKit React Native. The project explicitly states that graphical desktops and autonomous checkout are future work, not currently supported.
The project maintains active development with regular updates to its demo materials and documentation. The team provides detailed recording notes and reproduction steps for understanding the current interface behavior, including specifics about input controls and agent response handling. Documentation covers verification status of features, a roadmap for future capabilities, and architectural details. The codebase is available under an open license for inspection and modification, with contribution guidelines provided for those building on the foundation.
