Anthropic Puts Claude Cowork on Your Phone: Control Desktop AI Agents From Anywhere

Anthropic just handed knowledge workers a new way to boss around their AI. On July 7, the company updated its Claude mobile apps so subscribers can direct Claude Cowork, its desktop agent, straight from an iPhone or Android device. The move caps months of rapid iteration that turned a Mac-only research preview into a cross-device tool for getting real work done without sitting at a desk.

Claude Cowork first appeared in January as an offshoot of Anthropic’s coding agent. It gave regular users a way to ask Claude to handle tasks on their computers. By March the company had added computer-use abilities. Now Claude can see the screen, move the mouse, click buttons and type. Engadget reported the latest step. Update the Claude app on Android or iOS. A new Cowork tab appears in the sidebar. From there users monitor what the agent does back on the desktop machine at home or in the office.

But this isn’t just remote viewing. Anthropic made Cowork run tasks in the background. Earlier versions demanded a stable internet connection the whole time. No longer. The agent keeps going even if the connection drops. And when it needs permission for a sensitive step, a notification pops up on the phone. Approve or deny. “Nothing ships until you’ve reviewed and approved it,” Anthropic told Engadget.

The feature builds on Dispatch, introduced in March as a research preview for Pro and Max users. A persistent thread stretches between phone and desktop. Start a task on the train. Let Claude grind through files, spreadsheets or presentations on the office PC. Check progress or give final sign-off later. Forbes described the phone as a lightweight command surface. The heavy lifting stays local on the user’s machine. That choice limits cloud exposure but requires the desktop app to stay awake.

Practical examples pile up. Create a report pulled from local folders. Fill an Excel sheet with data scraped from multiple sources. Export a pitch deck as PDF ready for the next meeting. Claude seeks connectors such as Gmail, Google Drive or Slack to get the job done. CIO Dive noted these capabilities arrived under a research preview. The system runs locally on the device, much like OpenClaw, the agentic tool that grabbed attention earlier this year by letting users command computers through messaging apps.

Security questions follow close behind. Claude takes screenshots to understand the screen and decide its next move. The model avoids certain actions. It won’t trade stocks or process facial images. Anthropic advises users to close any windows with medical records, financial data or other sensitive material before handing over control. Similar warnings appear around OpenClaw. Only one in five companies planning agent deployments in the next two years have a mature governance model, according to a Deloitte study cited by CIO Dive.

Yet demand keeps growing. Nvidia CEO Jensen Huang called agentic systems the next big thing after ChatGPT. He urged every company to develop an OpenClaw strategy. Anthropic’s approach differs. It ties the agent tightly to the user’s own desktop rather than a fully cloud-based executor. That design appeals to enterprises wary of sending proprietary data off-site. But it also creates friction. The desktop must remain powered on and connected.

Anthropic’s release notes show steady progress. Cowork reached general availability on macOS and Windows in April. Computer-use preview landed March 23 alongside Dispatch improvements. Interactive apps arrived for iOS and Android on March 25. By July the mobile control layer had matured enough for wider rollout. Max subscribers receive access first. Other paid plans follow in the coming weeks.

Voice features expand the experience too. The Claude apps on Google Play and the Apple App Store tout dictation in multiple languages. No typing needed for brainstorming or working through ideas on the go. Analyze PDFs or screenshots directly from the phone. Generate SVG graphics or offer feedback on app designs. These additions make the mobile client more than a thin remote for the desktop agent.

Future plans point toward tighter integration. Anthropic intends to merge Cowork with the main chatbot. One interface for conversation and computer tasks. The change starts on web and desktop clients. Projects and artifacts will combine as well. Projects group chats and files around a theme to better use the model’s context window. Artifacts are the small apps and games Claude can build. Bringing them together could raise their profile. Few users engage with them today despite easier sharing.

Industry watchers see bigger forces at work. Agentic AI moves beyond chat into the background layer of daily work. Users assign recurring jobs. Daily email scans. Weekly reports. Claude runs them on schedule. The phone becomes the dispatch point. The computer does the execution. Results wait when the user returns. This pattern echoes what some have tested for months with earlier tools. Yet Anthropic packages it inside a single, governed product.

Challenges remain. Reliability varies with screen complexity. The model still scores below human levels on benchmarks like OSWorld. Self-correction helps but isn’t perfect. Enterprise adoption will hinge on governance. How do teams audit what the agent touched? Who approves recurring automations? Anthropic’s emphasis on review and approval offers one answer. Local execution offers another. Still, screenshots and file access create surfaces for error or misuse.

Even so, the trajectory looks clear. Mobile control of desktop agents lowers the barrier. Professionals no longer need to babysit long-running tasks. They start them from the airport, approve steps from a coffee shop, collect finished deliverables without opening a laptop. For teams already deep in Claude’s ecosystem the update feels less like a feature and more like the logical next layer. The AI doesn’t just answer questions. It acts on the user’s actual workspace. And now it does so under the user’s thumb from anywhere the phone has a signal.

Recent coverage reinforces the momentum. TechCrunch tracked the app’s rise to the top of the App Store earlier this year amid other news. Similarweb data showed Claude as the fastest-growing generative AI tool in early 2026, with mobile usage climbing alongside desktop. YouTube creators demonstrated remote control setups using QR codes for Claude Code, another related product. The pattern holds. Users want agents that work while they do other things. Anthropic is giving them exactly that, one careful update at a time.