Claude is starting to move beyond the chat box. Anthropic has introduced a new feature that lets its AI assistant use a person’s computer to complete tasks, including opening apps, browsing the web, and working with spreadsheets. It is a clear step toward AI tools that do more than answer questions.

The idea is simple, but the implications are not. Instead of only telling users what to do, Claude can now handle part of the work itself. That puts Anthropic more firmly into the race to build AI agents that can act on a user’s behalf, even when that user is away from the desk.

What Anthropic just announced

Anthropic announced Claude can now carry out tasks on a user’s computer. That includes opening apps, using a web browser, and filling in spreadsheets. The company announced the feature on Monday and framed it as part of a broader push to make Claude more useful as an AI agent, not just a chatbot that replies with text.

This is the center of the story. Claude is moving beyond conversation and into action. Instead of only telling you how to do something, it can now do part of that work on your machine after you give it a prompt.

Anthropic wants Claude to function as a tool that can handle parts of computer work while the user is away from the desk. That is the direction this release points to, and it is why the update stands out.

Claude’s New Computer-Use Release, Explained

Anthropic released a new feature that lets Claude use a person’s computer to complete tasks. Users can send Claude a request from a phone, and Claude can then act on the computer itself. That is a meaningful shift from the usual chatbot model, where the system stays inside a text box and leaves the real work to the user.

The company is not presenting this as a finished product for everyone. It is a research preview, which matters. That means Anthropic believes the feature is ready for real-world testing, but it also means the system is still early and likely to have limits and mistakes.

For readers who do not follow AI closely, the simplest way to understand this is that Claude is beginning to move from answering questions to carrying out tasks.

What Claude Can Actually Do on Your Computer

Claude can interact with a computer in ways that look familiar to any normal user. It can open apps, move through websites, click, scroll, and type. That matters because the tool is not limited to one screen or one narrow function. It is being positioned as something that can work across common computer actions.

One of the clearest examples is a user running late for a meeting and asking Claude to export a pitch deck as a PDF and attach it to a meeting invite. Claude can also fill in spreadsheets and work with files already on the computer. Those examples make the feature easier to understand because they reflect ordinary desk work.

Anthropic also says Claude can use connectors with apps such as Slack and Google Calendar. If a connector is available, Claude can use that route to complete the task. If not, it can still try to do the work by using the keyboard and mouse and moving through the computer manually, more like a human user.

How the feature works in real life

The basic setup is easy to grasp. A user gives Claude a task, sometimes from a phone, and Claude carries it out on the computer. That means a person does not always need to be sitting in front of the screen to start the work.

In practical terms, this could mean sending a request while commuting, stepping out, or moving between meetings. Instead of waiting until you get back to your desk, you assign the task and let Claude handle the steps. The meeting-file example shows the kind of office work Anthropic appears to be targeting first.

This feature also connects to Dispatch, which Anthropic released recently. Dispatch allows users to keep a continuous conversation with Claude from a phone or desktop and assign tasks remotely. Together, Dispatch and computer use push Claude closer to the role of a remote digital assistant rather than a simple chatbot.

Why Claude’s Computer Control Matters Beyond This Release

This matters because AI companies are trying to move beyond chat. For the past few years, many people have used AI mainly to write, summarize, or answer questions. Now the focus is shifting toward systems that can actually take actions. That is a different promise. It is less about giving advice and more about handling part of the work.

Anthropic’s move also fits into a broader trend around AI agents. OpenClaw, an open source framework, has played a major role in drawing attention to this category. Its rise earlier this year helped bring agent-style AI into the center of the conversation. In that context, Claude’s new feature looks like Anthropic’s response to a fast-moving change in the market.

There is also a competitive layer to the story. Nvidia has introduced NemoClaw, and other major tech companies, including OpenAI and Meta, are also working on agentic AI. Readers do not need to follow every company closely to see the bigger point: AI firms want their tools to do more than talk.

What Anthropic says about safety

Anthropic is not presenting this as risk-free. One of the main safeguards is that Claude will ask for permission before accessing new apps. That matters because giving software the ability to act on a computer raises obvious concerns about privacy, control, and actions a user did not intend.

The company also says the feature is still early and that Claude can make mistakes. That warning is important. A tool that can click, type, and move through apps may be useful, but it can also create problems faster than a text-only chatbot if it misunderstands the task.

Anthropic says it has added safeguards meant to reduce risks such as prompt injection. At the same time, it warns users not to rely on the feature for sensitive apps, and some of those apps are disabled by default. The message is clear enough: the company wants people to try the system, but with caution.

Who can use it right now

The feature is limited to paid Claude users. Access is available on Claude Pro, Max, Team, and Enterprise plans, which shows Anthropic is keeping the rollout inside its paid products rather than opening it to all users at once.

There is also a platform limit. Claude’s computer-use feature currently works only on macOS through the Claude desktop app. Users on Windows or other systems should not expect access yet.

For now, Anthropic is keeping the release fairly narrow. The company is giving access to paying users on selected plans while testing how the feature performs in real use.

FAQs

What is Claude’s new computer-use feature?

Claude’s new computer-use feature lets the AI do more than reply in a chat. It can now interact with a computer to handle certain tasks for the user, such as opening apps, moving through web pages, and working with files. That is the part that makes this update stand out. Claude is no longer being framed only as a writing or question-answering tool. Anthropic is pushing it toward something closer to a digital assistant that can actually carry out steps on a machine after receiving a prompt.

What kinds of tasks can Claude handle on a computer?

Claude can handle a range of desk-based tasks that normally require a person to click through apps and web pages. That includes opening programs, browsing websites, typing, scrolling, filling spreadsheets, and working with files. One example shown by Anthropic involves exporting a pitch deck as a PDF and attaching it to a meeting invite. The point is not that Claude can do everything, because it clearly cannot yet. The point is that it can now take over parts of routine computer work instead of only explaining how to do them.

Do I need to sit at my desk to use this feature?

No, that is one of the main ideas behind the release. Claude can be given a task remotely, including from a phone, and then carry it out on the computer. That changes the experience quite a bit. Instead of waiting until you are back at your desk, you can assign a job while moving between meetings, commuting, or stepping away for a while. Anthropic is clearly trying to make Claude feel less like a chatbot you consult and more like an assistant that can keep working while you are busy elsewhere.

Who can use Claude’s computer-use feature right now?

Right now, the feature is limited to paid Claude users. It is available to people on Claude Pro, Max, Team, and Enterprise plans. There is also an important device limit. At the moment, the feature works only on macOS through the Claude desktop app. So even if someone has a paid plan, they still need the right setup to use it. This is not a full open rollout. Anthropic is keeping access fairly controlled while it tests how the feature performs and how people actually use it.

Is Claude’s computer-use feature safe to use?

Anthropic is treating safety as a major part of the release, but that does not mean the feature is risk-free. Claude asks for permission before accessing new apps, which gives users a level of control. Anthropic also says the system is still early and can make mistakes, and some sensitive apps are disabled by default. That is a sensible warning. A tool that can click, type, and move through your computer can save time, but it can also cause trouble quickly if it misunderstands the task or runs into something unexpected.

Origianl Creator: Paulo Palma
Original Link: https://justainews.com/companies/anthropic/anthropic-releases-claude-feature-that-can-use-your-computer/
Originally Posted: Wed, 25 Mar 2026 18:39:31 +0000