KORA Challenges Windows and macOS With an AI-Agent-First PC OS

The team behind it has experience with Android, the Cyanogen custom smartphone image, and the Linux-based gaming OS GameOS.

Local inference and workspace isolation

The OS includes a hybrid inference engine called Cortex. It runs local AI models across the device's CPU, GPU and NPU, and can also draw on compute from other devices on the same network. Instead of organizing content by folder and app, KORA searches by meaning and builds the environment around task context and memory.

For security, KORA uses project-based "neural workspaces" with a permission system. An agent can use the data in its assigned workspace but cannot reach applications, agents or key information in others. The neural file system indexes only files inside the designated workspace, and permissions govern model and app access whether the model runs locally or in the cloud.

The company says it is taking an open approach, not tying users to any one vendor's models, and plans an SDK so outside developers can connect memory, projects and semantic search to their own services.

Hardware and open questions

The first target is the Qualcomm Snapdragon X2 Elite. KORA says it worked with Qualcomm for about a year to run a full Linux desktop on that chip with AI inference across CPU, GPU and NPU. Support will extend to Snapdragon X Elite, while Intel and AMD PCs have no schedule yet. Compatibility technology for running some Windows applications is also in development.

Challenges include security, privacy, accountability for agent errors, compatibility with Windows software and peripherals, and building a developer ecosystem. KORA Systems says pricing and the business model will be disclosed at the 2027 launch.