Close Your Laptops! Anthropic Launches Claude Cowork Agent for Mobile Devices

The time of needing to partially open a laptop to keep AI agents active is coming to an end. On Tuesday, Anthropic revealed that Claude Cowork, its AI agent designed to handle digital tasks on your behalf, is evolving beyond just the desktop application.

You no longer have to keep your device active, meaning you don’t need to leave your laptop open to maintain the agent’s function and run scheduled tasks overnight. Anthropic also announced limited versions of Cowork for users to engage with via its existing Claude smartphone app or a web browser, eliminating the prior need for a desktop connection.

In the promotional video for this new feature, Anthropic demonstrates a user requesting assistance with a business deal renewal planned for the next day. With a single prompt, the user asks Cowork to gather information from email threads, Slack channels, meeting transcripts, and recent online discussions. Subsequently, the user instructs Cowork to use that information to create a reference document for the meeting and a pre-composed email. Previously, Cowork could only accomplish this while your desktop session was active, but now, it can function even after you log off, capturing late-night incoming messages.

I first experienced Claude Cowork when it launched in January, and I was immediately impressed by its capability to fulfill the tasks I assigned it on my laptop, such as organizing a chaotic array of screenshot files into neat, labeled folders. It also performed well in helping me schedule events on my calendar. Although the agent wasn’t flawless and presented risks of prompt injections or other security vulnerabilities, Cowork represented a significant advancement in how everyday users could engage with their devices.

This isn’t the first instance of Claude users being able to interact with Anthropic’s agents on mobile devices. Previously, users could connect their smartphone app with their desktop through the Dispatch feature, allowing them to send task requests from anywhere. However, this method had a major drawback: “Your computer must be awake, and the app must be open for Claude to work on tasks,” according to Anthropic’s description. This is why some users kept their laptops open to maintain session activity. Now, Cowork is capable of executing tasks without an ongoing desktop session.

This announcement is part of a broader, recent trend in Silicon Valley toward continuously active, semi-autonomous AI agents that can be managed via text. The movement began with OpenClaw, a homebrew agent featuring a lobster mascot that went viral at the start of 2026, as early adopters utilized it around the clock and ceded control of their online activities.

Other tech companies envied the acclaim directed at this crustacean-themed agent. Consequently, during the first half of the year, OpenAI recruited OpenClaw’s creator and introduced Codex, its adaptive agent; Google unveiled Spark, its version of the always-on agent; and Anthropic focused on enhancing its agents’ user-friendliness. Anthropic’s standout offering was Claude Code, designed to help developers automate tasks. Cowork adopts a similar approach, extends it beyond the computer terminal context, and places that capability into chatbot form for average users.

Anthropic intends to launch this updated version of Cowork as a beta for subscribers of its Max plan, starting at $100 per month. The features are expected to gradually become available to members of Anthropic’s more affordable Pro tier, priced at $20 per month. It remains uncertain whether this will be extended to free users, who currently do not have access to Claude Cowork in their subscription tier.