Anthropic is taking Claude Cowork out of the desktop app and into the places where people actually spend their day—phones and browsers. Starting Tuesday, the company’s AI-powered “work sessions” platform will be available on mobile and web for the first time, marking a meaningful shift in how Cowork is meant to be used. Until now, Claude Cowork has been largely tied to the Claude desktop experience on macOS and Windows. With this expansion, iOS and Android users can finally access Cowork as well, and web users get a new way to start and continue sessions without installing or opening a dedicated desktop client.
The timing also matters. Anthropic says the rollout begins with Max subscribers, then expands to other Claude plans “in the coming weeks.” That staged approach suggests the company is balancing demand with infrastructure and safety considerations—especially because Cowork sessions are not just chat prompts. They’re designed to run work-like flows that can involve tools, context, and longer-running tasks. Moving those sessions into more environments than before—mobile operating systems and a browser runtime—adds complexity, and it’s reasonable to expect Anthropic to ramp access while it monitors performance and user behavior.
At the center of the update is a simple but important idea: Cowork should be portable. In practice, that means you shouldn’t have to wait until you’re at your desk to begin a task, hand it off, or pick it up again later. Anthropic says Cowork sessions will now run in the cloud by default. That change is what makes cross-device continuity feel natural rather than fragile. Instead of treating Cowork like something that only exists inside a specific app instance on a specific machine, the platform can treat a session as something you can return to—whether you started it on a phone, continued it on a laptop, or resumed it in a browser tab.
This is a subtle shift in product philosophy. Desktop-first AI tools often assume that the “real work” happens on a computer, and the rest of the day is just waiting. But modern workflows don’t behave that way. People capture ideas on mobile, review documents on the go, and do quick iterations between meetings. If Cowork is meant to help with real tasks—planning, drafting, coding assistance, organizing information, and coordinating steps—then limiting it to desktop is a bottleneck. By expanding access, Anthropic is effectively reducing the friction between “I have an idea” and “I can start a structured AI work session.”
Still, Anthropic is careful not to oversell parity. The company says the “full experience” for Cowork remains on the desktop app, including features like local file access. That distinction is crucial for anyone who has tried to use AI tools across devices: the moment you need deep integration with your files, your filesystem, or your local environment, the browser and mobile apps can’t always replicate the same capabilities. Local file access isn’t just a convenience; it’s often the difference between an AI assistant that can talk about your project and one that can actually operate on it directly.
So what does “full experience” mean in this context? It likely refers to the kinds of workflows where Cowork can read from and write to files stored on your device, interact with local resources, and maintain a tighter loop between the AI and your working environment. On mobile and web, those capabilities may be limited by platform permissions, sandboxing, or the absence of direct filesystem access. Anthropic’s message implies that while you can use Cowork on mobile and web, the most powerful version—especially for workflows that depend on local assets—still lives in the desktop app.
That said, the cloud-by-default change suggests that many of the core benefits of Cowork can still carry over. If sessions run in the cloud, then the “brain” of the work session doesn’t have to be trapped inside a single device. You can imagine a workflow like this: you start a Cowork session on your phone during downtime, give it a goal and relevant context, and let it progress. Later, when you’re back at your laptop, you can open the desktop app and continue with the richer toolset, including local file access. Or you might start in a browser at work, then finish at home on desktop. The point is not that every feature is identical everywhere—it’s that the session itself can persist.
There’s also a broader implication here for how AI agents are expected to behave. Cowork is positioned as a platform for running AI-powered work sessions, which is different from the typical “one prompt, one response” pattern. When products move toward agentic workflows, the user experience becomes less about typing and more about managing state: what’s been done, what’s next, what context is available, and where the work is stored. Cloud execution by default is one of the most practical ways to make that state manageable across devices.
In other words, Anthropic isn’t just adding buttons for mobile and web. It’s aligning Cowork with the operational reality of agent workflows: they need continuity, and continuity is easiest when the session state is centralized. That’s why the cloud default is such a big deal. Without it, cross-device use tends to degrade into awkward workarounds—copying text, re-uploading context, or restarting tasks from scratch. With cloud sessions, the platform can preserve the thread of work more reliably.
For users, the immediate benefit is convenience. For teams and power users, the benefit is workflow design. Once Cowork is accessible from a browser, it becomes easier to integrate into daily routines that already revolve around web apps. A browser-based entry point can also reduce the “activation energy” required to start a session. Instead of thinking, “I need to open the desktop app,” you can think, “I can start Cowork right here.” That matters for adoption, because the best AI tools are the ones that fit into existing habits rather than demanding new ones.
Mobile access changes the same equation in a different direction. On iOS and Android, the biggest challenge for AI tools is often not intelligence—it’s usability. Typing long instructions on a phone is painful, and managing multi-step tasks can feel clunky. But Cowork’s session model can help. If you can initiate a work session with a clear objective and then let it run, you don’t have to stay glued to the screen. You can provide initial direction, check back later, and then refine when you have better input capacity. That’s a more realistic model for mobile usage than expecting users to conduct complex back-and-forth conversations entirely on a small device.
Of course, there are tradeoffs. Mobile environments are constrained by screen size, input methods, and app permissions. Web environments are constrained by browser security models and the lack of direct access to local files unless the user explicitly grants it through supported mechanisms. Anthropic’s statement about local file access being part of the desktop “full experience” hints that these constraints will shape what users can do on mobile and web. But even with limitations, the ability to start and continue sessions can still deliver substantial value—especially for tasks that don’t require deep local integration.
It’s also worth considering what this expansion signals about Anthropic’s competitive posture. Many AI platforms are racing to become the default interface for AI assistance. Chatbots are easy to access, but they often fall short when users want structured, multi-step outcomes. Agentic tools promise more, but they also risk becoming harder to use if they’re locked behind a single platform. By bringing Cowork to mobile and web, Anthropic is making it harder for competitors to frame Cowork as “desktop-only” or “not for everyday use.” The more places Cowork can appear, the more likely it is to become part of routine workflows.
There’s another angle: the rollout starting with Max subscribers suggests Anthropic is testing demand and load under real-world conditions. Cloud execution by default means sessions may consume compute resources continuously or in bursts depending on the task. Scaling that across more devices and more users requires careful monitoring. A staged rollout helps ensure that the experience remains stable as the platform expands beyond the desktop audience that previously had access.
For existing desktop users, the update may feel like a quality-of-life improvement. For new users, it could be the difference between curiosity and adoption. People who haven’t tried Cowork because they didn’t want to install or open a desktop app now have a lower barrier. And once they experience the session model—where work can continue beyond a single interaction—they may be more inclined to explore the desktop app later for the full feature set.
The “coming weeks” language for other Claude plans also implies that Anthropic expects a gradual expansion rather than a single day of universal access. That’s common for major feature rollouts, but it also creates a window for early adopters to learn the workflow. In the early phase, users who get access first can develop best practices: how to start sessions effectively on mobile, what kinds of tasks work best in the browser, and how to transition between devices without losing momentum. Those patterns often become the unofficial playbook for how a new AI feature is used.
From a user perspective, the most interesting part of this update is how it changes the rhythm of work. Traditionally, AI assistance is something you consult when you’re stuck. Cowork’s session approach encourages a different rhythm: you set a goal, let the system run a work process, and then return to review and steer. When that process can be initiated from anywhere, the “stuck moment” becomes less important than the “start moment.” You can begin work earlier, even if you can’t fully execute it immediately.
Imagine a scenario: you’re away from your computer but you want to kick off a research or planning task. With mobile access, you can start Cowork, provide the objective, and perhaps outline constraints. Later, when you’re at your desk, you can review what the session produced and decide whether to continue, refine, or pivot. This reduces the latency between intention and action. It also makes AI assistance feel less like a tool