Kora OS: An AI-First Alternative to Windows
Kora plans a free Linux desktop built around AI models and agents. See how it could challenge Windows, where its business model fits, and what remains unproven.

Kora Wants to Make the Desktop an Operating System for AI
Most AI tools still live inside a browser tab or a single app. That's fine for drafting an email or summarizing a document. It gets awkward when an agent needs to find a file, keep context as you bounce between tasks, or work across multiple apps without losing what you asked it to do.
Kora Systems wants to tackle that problem at the operating-system level. The company, co-founded by former Cyanogen CEO Kirt McMaster, has announced a Linux-based desktop meant to spread AI models and agents across the whole computing experience. Kora says it's targeting a 2027 launch as a free-to-use operating system for Intel-, AMD- and Qualcomm-powered PCs.
It's an ambitious target. Windows has a huge lead, and an AI-friendly desktop can't just look impressive - it has to be reliable and secure, play nicely with everyday software, and still feel worth paying for months (and years) in. This article breaks down what Kora is building, why McMaster thinks an operating system is the right place for AI, and what the company still needs to prove.
What is Kora OS, and what will it do?
Kora OS is a planned Linux desktop operating system designed to help AI models and agents work across applications, files and tasks. The idea is simple: instead of treating AI like a feature inside a single app, Kora wants it to be part of the desktop itself.
In practice, Kora's goals include shared context between tasks, access to files, and separate workspaces where agents can operate with restricted access to resources. The company suggests this could let a user ask an agent to find relevant documents, pull information from multiple apps, and keep going on the same job without manually recreating the context each time.
Those are goals described by the company, not proof that every workflow already works reliably. That distinction matters. An operating system can make integration easier - but it also has to handle permissions, application compatibility, updates, and failures across the whole machine.
From Playtron to an AI desktop
Kora was previously known as Playtron, a company working on a Linux-based gaming operating system called GameOS. In July 2025, it rebranded as it shifted toward AI. Kora says its new work builds on nearly three years of GameOS development instead of starting from scratch.
That base could help with some of the heavy lifting, but gaming and AI workflows put different demands on a desktop. GameOS gave the team experience with Linux and consumer hardware. Kora still has to show that its new interface and AI services handle day-to-day work better than what people already rely on.
Kora says the software is based on Linux and uses Nix. It also says it moved away from the Fedora Silverblue foundation used by GameOS. Linux and Nix are open source, but McMaster says Kora's Cortex AI engine and desktop interface are proprietary.
Hardware and availability
Kora expects to support PCs with Intel and AMD processors, plus Qualcomm Snapdragon-based machines. The company says it has worked with Qualcomm engineers for nearly a year on optimization for Snapdragon-powered PCs, including systems with Snapdragon X2 Elite chips.
Kora is expected to launch in 2027. It isn't available as a mainstream desktop operating system today, and Kora hasn't named PC manufacturing partners or shared a clear partner timeline. Those details will matter, because preinstallation and dependable hardware support are usually the difference between "interesting" and "easy to try."
Why is Kora challenging Windows?
Kora isn't trying to convince every Windows user to switch. McMaster says the company is aiming at people who use AI heavily and are willing to pay for models and services, rather than trying to win over Windows' entire audience.
The market share gap is big. Statcounter's web traffic data puts Windows at about 77% of the worldwide desktop operating-system market in October 2026, compared with nearly 10% for Apple's OS X. A new operating system has to overcome more than brand recognition - people rely on familiar applications, workplace support, peripheral compatibility, and years of habits.
Kora's narrower focus is a practical response. If the product is built for a specific group with demanding AI workflows, it doesn't have to replace Windows everywhere to start gaining users. Still, a focused market isn't automatically a viable business. Kora will need to show that these users have recurring needs its system meets significantly better.
The lesson from Cyanogen
Before Kora, McMaster led Cyanogen, which tried to loosen Google's grip in Android. Cyanogen offered an alternative operating-system experience, but it stayed heavily tied to Google Mobile Services. McMaster says that dependency limited the company's ability to build a truly independent ecosystem.
His stated takeaway is that a challenger can't rely too much on the incumbent it's trying to displace. That matters for Kora, because an AI desktop depends on more than the interface - models, cloud services, hardware support, and applications all shape the experience.
Kora says it wants to support models and services from multiple providers. That could reduce dependence on any single AI company, but the company hasn't committed to open-sourcing its own software. Where it lands on openness vs. proprietary control will affect how much flexibility users and developers actually get.
Why not build an AI browser?
Kora's shift toward an operating system came as companies like Perplexity and OpenAI pushed deeper into AI-powered browsers. McMaster argues that a browser doesn't give agents enough access to complete more complex computer tasks.
There's a clear product difference. Browsers can reach web content and browser-based apps, but a lot of real desktop work involves local files and native software. In theory, an operating-system-level agent can cross those boundaries.
The trade-off is that more access comes with more responsibility. It can make agents more useful, but it also increases the stakes when something goes wrong - like a bad instruction, a compromised model, or permissions that are too broad. Kora will need to make its access controls understandable and enforceable, not just technically possible.
How does Kora plan to integrate AI across the desktop?
Kora's core design is to make models and agents available across applications, with context that can persist between tasks. The company also describes separate workspaces where agents can operate with restricted access to files and other resources.
This addresses a real limitation of app-by-app AI: when each tool has its own conversation history and file access, users keep moving information between services. A desktop layer could reduce that friction. But the operating system also has to make it obvious what the agent can see, what it changed, and when it's acting on the user's behalf.
Local and cloud processing
Kora says its Cortex engine is designed to run AI tasks locally when possible, and use cloud-based models when more compute is needed. A hybrid setup can balance responsiveness, capability, and cost. Local processing can also keep some information on the device, while cloud models handle tasks that exceed local hardware.
Either way, users still need clear data handling. They should know which tasks leave the machine, which model processes them, and whether files or conversation history are retained. For businesses, these questions connect directly to confidentiality, compliance, and internal security rules.
Kora's source announcement doesn't say which models will be available at launch, how users will choose between them, or how local-vs-cloud routing will work in detail. Those aren't minor implementation notes - they shape whether users can control costs and data exposure or have to rely on opaque defaults.
Permissions are part of the product
An agent that can access files and operate across apps needs boundaries. Kora's restricted workspaces hint at a permission model, but the real question is how granular and visible it will be.
A good implementation should make it clear:
- Which files and applications an agent can access.
- Whether access is temporary or persists across tasks.
- What actions require confirmation before they happen.
- How a user can review or undo changes.
- Which information is sent to a cloud service.
This isn't "nice to have" polish. It's core to safety and usability. If users can't reasonably predict what an agent will do, they'll either avoid it altogether or grant broad access and accept risks they don't fully understand.
AI-assisted development is not product validation
Kora has a team of about 50 engineers and says it uses coding tools including OpenAI's Codex and Anthropic's Claude. McMaster told TechCrunch that these tools accelerated development by at least fivefold, with much of the improvement coming in the previous six months.
That could help a smaller team build and iterate faster. But it doesn't, by itself, prove the system is secure and stable enough for a broader user base. AI coding tools can speed up writing code - they don't eliminate the need for testing, code review, hardware validation, and long-term maintenance.
The real metric isn't how quickly a team can ship. It's whether Kora can deliver a dependable desktop, support it across different hardware setups, and keep its AI features useful as models and services change.
Can Kora build a viable business beyond the operating system?
Kora plans to make the desktop operating system free to use, then earn revenue through subscriptions that bundle access to AI models and agentic services. That structure separates the cost of installing the system from the cost of using more capable AI.
It also lowers the barrier to trying Kora. On the other hand, it ties the company's revenue to the economics of AI inference - cloud models cost money to run, and subscriptions have to cover those costs while still feeling attractive to users. Local execution could reduce costs for some tasks, but Kora hasn't disclosed pricing or the specifics of its subscription tiers.
McMaster estimates that 45 million to 50 million people already account for a large share of AI usage and revenue, and he expects that market to exceed 200 million users by 2033. Those are his estimates, not independently verified forecasts. The business question is whether a meaningful share of those users actually need a separate operating system - rather than just using AI services layered onto Windows or macOS.
Enterprise and hardware partnerships
Kora is working with Saudi Arabia's AI company Humain on an enterprise version called Humain OS, expected to run on Snapdragon-powered laptops in 2027. A dedicated enterprise product could be a route into organizations that want managed AI workflows. But it also brings higher expectations around deployment, administration, security, and support.
Kora says it plans to announce PC manufacturer partnerships later, but it hasn't named companies or shared a timeline. That's a key uncertainty. Operating systems benefit from distribution, and device makers need confidence that the platform will keep working for their customers after purchase.
Kora has also said gaming remains part of its plan. It intends to build AI-powered tools for creating, publishing and playing games within the operating system, and Android app support is on the roadmap. Those moves could broaden the product's appeal, but they also mean more engineering and compatibility work. In the early days, a focused first release may matter more than a long list of future features.
Funding, independence and execution risk
Kora's investors include Samsung Next, Square Enix, Alumni Ventures, Polychain, VanEck and Circle - all of which previously backed Playtron. McMaster says the company has raised additional funding since shifting to AI and has enough runway to operate for several years. He didn't disclose the total raised or the company's valuation.
Funding can give the team time to build - but it doesn't solve the platform challenge on its own. Kora needs a clear reason for users to switch, a way to support their existing workflows, and recurring revenue that can pay for AI services without making the product's costs unpredictable.
The best early sign won't just be curiosity. It will be sustained use: people choosing Kora for real work, coming back to its agents, and paying for services because the operating-system integration actually saves time or effort enough to justify moving.
Conclusion
Kora's pitch is straightforward: build an operating system for people who want AI agents to work across their desktop, instead of keeping AI trapped inside isolated apps. Linux, a planned multi-model approach, local processing, and restricted agent workspaces give the company a path to try that idea.
The hard part is turning that design into a dependable product. Kora still has to prove its permissions are understandable, that the software works in real-world workflows, that hardware support is broad enough, and that the subscription economics make sense. Kora doesn't need to replace Windows' mainstream user base to find a market - but it does need to give its target users a concrete reason to switch.
For now, it's best to treat Kora as a credible direction, not a proven Windows replacement. The real test will be a usable release and evidence that people rely on it for work they can't handle as effectively with their current setup.