Kora Wants to Rebuild the Desktop Around AI Agents

Windows has owned the desktop for years, but Kirt McMaster wants to change what a desktop operating system is built to do. His company, Kora Systems, has unveiled a Linux-based platform designed for people who rely on artificial intelligence throughout their everyday work.

Kora is not positioning itself as a replacement for every Windows user. It is aiming at “the AI native, not mass market,” with a system that brings models, autonomous agents, files, applications, and user context together at the operating-system level. The software is scheduled to launch sometime in 2027 as a free-to-use desktop operating system.

An Operating System Built Around AI Agents

Today’s AI tools often live inside individual applications or separate browser tabs. Kora takes a different approach: its AI models and autonomous agents can work across applications, access files, and retain context between tasks. The system can track conversations, documents, and other information across multiple AI tools, helping power users move between providers without losing the thread.

Agents can also operate inside separate workspaces with restricted access to files and other resources. That design gives users a way to assign tasks while controlling what each agent can reach. Franck De Girolami, one of Kora’s three founders, described the idea in simple terms: “All your context, all your memory is on your machine,”.

Kora is a standalone operating system built on the Linux kernel, and it runs directly on hardware. The company is building support for AI models and services from different providers rather than tying the experience to one model company. Its proprietary technology includes the Cortex AI engine and the desktop interface, while Kora has not committed to making its own software open source.

The system will support x86-64 architecture, existing Intel- and AMD-powered PCs, and machines built around Qualcomm’s Snapdragon chips. Kora has worked with Qualcomm’s engineers for nearly a year to optimize the platform for Snapdragon-powered PCs, including Qualcomm’s Snapdragon X2 Elite chips. Kora has partnered with Qualcomm to bring the operating system to Snapdragon PCs next year.

Why Kora Is Not Chasing Every Windows User

That is a bold target in a market where Microsoft remains dominant. Windows held about 77% of the worldwide desktop operating-system market in October, based on StatCounter web traffic data, while Apple’s OS X held nearly 10%. Breaking into that market would challenge any new operating system, but McMaster says Kora does not need to win over Windows’ entire user base.

“Microsoft can keep its billion users,” McMaster said. Kora is targeting people who use AI extensively and are willing to pay for access to powerful models and services. McMaster estimates that 45 million to 50 million people currently account for much of AI usage and revenue, and he expects that group to grow beyond 200 million users by 2033.

The company is optimizing Kora for desktop configurations with at least 64 gigabytes of RAM, twice the memory capacity of the top Googlebook. That hardware focus reflects the demands of running AI models that need serious chip power, rather than treating AI as a small feature added to a familiar desktop.

Google launched its Googlebook OS and compatible laptops in October around Gemini AI, but McMaster does not see that operating system and its Android support as satisfying power users who want to run demanding AI models. Kora also plans built-in compatibility layers for standard Windows apps, while the company develops native apps for contacts, calendar, and email.

“We can evolve much faster than macOS and Windows,” McMaster said. That speed is central to Kora’s pitch: the company wants the operating system itself to change around the way people use AI, instead of adding AI tools to an older desktop structure.

From Gaming Platform to AI-Native Desktop

Kora grew out of Playtron, a gaming operating-system company led by McMaster that rebranded in July 2025 as its focus shifted to AI. The company had already spent nearly three years developing GameOS, a Linux-based gaming operating system primarily built for portable handhelds such as a Steam Deck, with PC support as well.

That existing architecture now lives inside Kora as GameFrame, allowing users to run PC games when their hardware can keep up. Gaming remains part of the company’s plans, and Kora intends to introduce AI-powered tools that let developers create, publish, and play games directly within the operating system. Support for Android apps also sits on the roadmap.

Kora uses Nix and replaces the Fedora Silverblue foundation of Playtron’s earlier GameOS. The company is also developing a software development kit for third-party developers, opening a path for applications built around Kora’s system-level AI features.

The company is being developed by about 50 engineers, and Kora relies on AI coding tools including OpenAI’s Codex and Anthropic’s Claude. McMaster said those tools have accelerated development by at least fivefold, with most of the gains arriving over the past six months.

McMaster’s earlier experience shaped the company’s independence strategy. He cofounded and helped lead Cyanogen, an open source Android fork that gained millions of users before shutting down around a decade ago. Cyanogen offered features beyond Google’s version of Android but remained heavily dependent on Google Mobile Services, limiting its ability to build an independent ecosystem.

“If you are trying to compete with an incumbent but are still reliant on them, you have no chance,” McMaster said. None of Kora’s engineers worked at Cyanogen; instead, the company recruited heavily from the Linux gaming community.

McMaster believes browsers cannot give AI agents enough access to a computer for complex tasks. Early companies including Perplexity and OpenAI were moving into AI-powered web browsers, but Kora chose the operating system as its starting point. Its next test will be whether an AI-native desktop can turn a focused group of power users into a much larger market by 2027.

“When you’re a billion-user incumbent, it’s impossible to tip the boat, man. It’s just really hard to make big changes. Over time, there’s this event horizon where building a new kind of OS for intelligence can become something very compelling.”

Stay connected via Google News

Follow us for the latest AI News