The era of half-cracking a laptop to keep AI agents running comes to a close. On Tuesday, Anthropic announced that Claude Cowork, its AI agent designed to perform digital tasks for you, is expanding beyond the desktop app.

You don’t need to leave your device active, aka laptop cracked open, to keep the agent going and run scheduled tasks overnight. Anthropic also announced limited versions of Cowork for users to interact with via its existing Claude smartphone app or the web browser, sans the formerly required desktop connection.

In the launch video for this feature, Anthropic shows someone asking for help with a business deal renewal scheduled for the following day. In a single prompt, the user asks Cowork to pull together data from email threads, Slack channels, meeting transcripts, and recent online chatter. Then the user asks Cowork to use that info to generate a reference document for the meeting and a prewritten email. Cowork could previously do all of that while your desktop session was active, at least. Now, the agent can run even after you clock out, catching those late-night incoming messages.

I first tried Claude Cowork when it dropped in January, and I was immediately stunned that this agent could actually follow through and complete the tasks I asked it to run on my laptop, like organizing a mess of screenshot files into respectable, labeled folders. It also did a solid job of helping me schedule events on my calendar. The agent wasn’t perfect, and it still exposed me to the risk of prompt injections or other security breaches, yet Cowork felt like a step change in how everyday users could interact with their devices.

This isn’t the first time Claude users have been able to interact with Anthropic’s agents on their mobile devices. Previously, users could pair their smartphone app with their desktop through the “Dispatch” feature. This enabled users to send task requests from their phone, no matter where they were. But this approach had one major limitation. “Your computer must be awake, and the app must be open for Claude to work on tasks,” reads Anthropic’s description. That’s why some users left their laptops cracked open to keep sessions running. Now, Cowork can run tasks without an active desktop session.

This announcement is part of a larger, recent shift in Silicon Valley towards always-running, semi-autonomous AI agents that you can control via texting. The trend was sparked by OpenClaw, a homebrew agent with a lobster mascot that went viral at the beginning of 2026, as early adopters ran it 24/7 and handed over control of their online lives.

Other tech companies were jealous of all this crustacean-focused praise. So, in the first half of the year, OpenAI hired OpenClaw’s creator and launched Codex, its adaptive agent; Google launched Spark, its take on the always-on agent; and Anthropic leaned further into making its agents more user-friendly. Anthropic’s breakout hit was Claude Code, which helped developers automate tasks. Cowork takes a similar approach, translates it out of the context of the computer terminal, and puts that power into chatbot form for average users.

Anthropic plans to roll out this revamped version of Cowork as a beta to subscribers of its Max plan, which starts at $100 a month. Then, the features are expected to trickle down to members of Anthropic’s cheaper tier, Pro, which costs $20 a month. It’s unclear whether this will roll out to free users, who don’t currently have access to Claude Cowork in their subscriber tier.