Anthropic is changing how users can run Claude Cowork, its AI agent for digital tasks. The company announced on Tuesday that Cowork is expanding beyond the desktop app, bringing limited versions into the existing Claude smartphone app and the web browser.
The practical shift is simple: users no longer need to keep a laptop open and active just to let the agent continue working. Cowork can now run tasks without the formerly required desktop connection, including scheduled work that might continue overnight.
What Anthropic Changed
Claude Cowork was already designed to carry out digital work on behalf of users. The previous version, however, depended on an active desktop session. If the computer was asleep or the app was closed, the agent could not keep working.
That limitation mattered because Cowork was meant to do more than answer questions. It could organize files, help schedule calendar events, and pull together materials across workplace tools when the desktop setup was available.
Anthropic’s new release changes the access model. Users can now interact with limited versions of Cowork through the Claude smartphone app or a web browser, without maintaining an active desktop connection.
The source article describes a launch video in which a user asks Cowork to prepare for a business deal renewal scheduled for the following day. In one prompt, the user asks the agent to gather information from email threads, Slack channels, meeting transcripts, and recent online chatter. Cowork is then asked to turn that material into a reference document and a prewritten email.
The important point is not just that Cowork can perform that kind of task. It is that the agent can keep running after the user has stepped away, including handling late-night incoming messages.
Why the Desktop Requirement Was a Problem
Claude users previously had a way to send agent tasks from a phone through Dispatch. That feature paired the smartphone app with the desktop, so users could issue requests while away from their computers.
But Dispatch still depended on the computer. Anthropic’s description said, “Your computer must be awake, and the app must be open for Claude to work on tasks.” That meant mobile access was not the same as mobile independence.
For users trying to run longer jobs, the restriction created a clumsy workflow. If the agent needed to keep going, the machine had to stay awake. The new Cowork approach removes that specific dependency for the revamped experience.
This makes Claude Cowork more consistent with how people already use chatbots on phones. Instead of treating the phone as a remote control for a desktop agent, Anthropic is moving the agent experience closer to the chat interface itself.
Where Cowork Fits in the Agent Race
The announcement lands during a broader shift in Silicon Valley toward always-running, semiautonomous AI agents that users can control through texting. The source article traces that trend to OpenClaw, a homebrew agent with a lobster mascot that went viral at the beginning of 2026.
OpenClaw drew attention because early adopters ran it 24/7 and gave it control over parts of their online lives. That attention helped set the direction for larger companies building agent products.
In the first half of the year, OpenAI hired OpenClaw’s creator and launched Codex, described as its adaptive agent. Google launched Spark, its own take on the always-on agent. Anthropic, meanwhile, continued making its agents easier for users to operate.
Anthropic already had Claude Code, which helped developers automate tasks. Cowork applies a similar idea outside the computer terminal. It puts agentic work into a chatbot-style experience aimed at average users rather than only developers.
Who Gets Access First
Anthropic plans to roll out the revamped Cowork as a beta for subscribers to its Max plan, which starts at $100 a month. After that, the features are expected to reach Pro users, whose plan costs $20 a month.
It is unclear whether the feature will become available to free users. The source article notes that free users do not have access to Claude Cowork in their subscriber tier.
Alongside the release, Anthropic also published a report on Claude Cowork usage patterns. The company says white-collar laborers are increasingly using its tools in their workflows.
The two largest recent usage categories named in the report are “Business process and operations,” including data reports and checklists, and “content creation and copywriting,” including slide decks and partnership proposals.
The Smartphone Becomes the Control Surface
Anthropic is not alone in trying to connect chatbots and agents more tightly on mobile devices. OpenAI rolled out Codex Remote in June, a feature similar to Claude Dispatch that lets users control desktop agents from smartphones.
OpenAI also launched more Codex-focused updates for its iOS app in July. Those updates included “support for creating, searching, opening, forking, and managing Codex tasks directly from a conversation.”
Anthropic’s new release goes further by merging the Claude chatbot interface and the Cowork agent for browser and desktop versions. That points to a larger product strategy: make agentic automation feel like a natural extension of the chatbot people already use.
The bigger implication is that AI agents are moving away from standalone tools and toward everyday interfaces. Instead of asking users to open a special app or maintain a desktop session, companies are trying to place agent controls inside the phone-based chatbot experience.
For Claude Cowork, that means the laptop is no longer the center of the workflow. The agent can be started and managed through more familiar surfaces, while continuing work without the user keeping a desktop session alive.