# Claude Computer Use Comes to Team and Enterprise Plans

> Anthropic put Claude computer use in beta for Team and Enterprise on Nov 20, off by default. Who can turn it on, app block lists, and the risks it names.

- Author: Nitish Garg, Founder & CEO, CellCog
- Published: 2026-11-21
- Canonical (HTML): https://cellcog.ai/blog/claude-computer-use-team-enterprise/
- Section: Guides / Choosing a platform
- Publisher: CellCog (https://cellcog.ai), the AI employee platform. Blog index for agents: https://cellcog.ai/blog/llms.txt

## Key points

- Anthropic put computer use in beta on Claude Team and Enterprise plans on November 20, 2026, rolling out to organizations gradually.
- It is off by default; only an Owner or Primary Owner can turn it on, under Organization settings, Capabilities.
- Admins can block chosen apps, such as a password manager, or allow only a list; the lists do not cover Claude in Chrome, websites or remote desktops.
- Anthropic says there is no sandbox between Claude and a user's apps, so content on screen could steer Claude into unintended actions.
- Claude's screenshots are saved with the chat, and organizations on HIPAA-ready, CMEK, Zero Data Retention or FERPA settings cannot turn it on.

## At a glance

- **What launched?** Computer use in beta for Team and Enterprise plans: Claude sees a user's screen and clicks and types in the desktop apps they allow.
- **Where does it run?** In the new Claude experience in Claude Desktop on macOS 14 or later and Windows, and from the web or mobile apps when Computer access is on.
- **Is it on by default?** No. An Owner or Primary Owner turns it on, and on Enterprise plans can keep it from chosen roles.

**Claude can now click and type in a company's desktop apps, if an admin lets it.** Anthropic's [release notes](https://support.claude.com/en/articles/12138966-release-notes) for November 20, 2026 list computer use in beta on Team and Enterprise plans, "rolling out to organizations gradually." Claude takes screenshots of a user's screen and works in the desktop apps that user allows.

The [setup article](https://support.claude.com/en/articles/18483812-set-up-computer-use-for-team-and-enterprise-plans) is mostly about control. Computer use is off by default, only an Owner or Primary Owner can turn it on, and Anthropic lists in detail what the admin settings do not reach.

## Where it works, and where it does not

*Table: Availability of Claude computer use, from Anthropic's setup article (November 2026)*

| Surface or setting | Computer use |
|---|---|
| Claude Desktop, new Claude experience, macOS 14+ or Windows | Yes, once the admin and the user turn it on |
| Claude on the web and mobile | Yes, if Computer access is on and Claude Desktop is open on the computer |
| Tasks started with the separate Cowork option | No |
| Claude Code | No |
| HIPAA-ready, Zero Data Retention for Claude Code, CMEK, FERPA or K-12 settings | Cannot be turned on |

It needs the Chat and Cowork unified capability. On a Mac, Claude Desktop also needs the Accessibility and Screen Recording permissions, which an MDM profile can prepare but cannot switch on for computer use.

## What admins control

- **The switch.** Organization settings, Capabilities, then Computer use under Claude Desktop. Turning it off takes about an hour, sometimes longer, to reach each user's app.
- **Who gets it.** On Enterprise plans it is a capability in custom roles. Team plans have no custom roles, so the setting covers every user.
- **Which apps.** App access takes a block list (a password manager, for example) or an allow list. Users still approve each allowed app. The list is sent to users' computers, and Anthropic warns not to include anything confidential.
- **Which computers.** A device-management key, computerUseEnabled set to false, keeps it off on a machine; it needs Claude Desktop 5.1231.0 or later. Anthropic is plain about its limits: "It isn't a lock against someone who can change the app or the computer's settings."

![Animation: an isometric platform with an admin panel whose switch starts off and slides on in teal; a monitor with three app windows where a teal cursor clicks in the first two; an orange plate with a padlock drops over the third window; then the line Off by default. No sandbox once it is on.](https://cellcog.ai/blog/media/claude-computer-use-team-enterprise/computer-use-switch.svg)
*An admin turns it on, Claude works in allowed apps, and a blocked app stays shut*

## What Anthropic says the controls do not cover

The setup article spends as much space on gaps as on settings. "Computer use has no sandbox between Claude and a user's applications. Content on the screen, such as web pages or emails, could cause Claude to take unintended actions." And: "Computer use isn't limited to the folders a user has connected." Folder limits and read-only folders do not apply to it.

*Table: What the app block and allow lists do not cover, per Anthropic*

| Not covered | What that means |
|---|---|
| Claude's other tools | Claude in Chrome, the built-in browser, Claude Code and desktop extensions |
| Websites and browser extensions | A password manager's extension in an approved browser, for example |
| Remote desktops and virtual machines | Apps inside an approved remote window |
| Text copied from a denied app | Claude can read it once pasted into an approved app |
| Record a skill | A recording includes every app on screen |

Screenshots are saved with the chat and can include confidential information. The automated check that reviews Claude's actions can see the latest screenshot, and Anthropic can keep a copy of it in its record of that check.

Anthropic's [Cowork safety guide](https://support.claude.com/en/articles/13364135-use-claude-cowork-safely) frames the risk the same way: "Isolation limits where Claude's code runs. It doesn't limit what Claude reads or does." Its advice is to limit what Claude can read from outside a trusted boundary, and what it is allowed to do.

## What it means for teams

Agents that operate a screen are arriving everywhere at once: Gemini on the Mac, ChatGPT in Chrome, and now Claude across a company's desktop apps. The question for an operations or IT owner is less whether the agent can click a button than who decided it could, and how fast that decision can be taken back. Anthropic's answer is an off switch per organization, roles, app lists and a device key, with an hour of delay on the way out.

CellCog's AI employees take a different route to the same work. They run on their own computer with their own files and browser identity; when a user connects their own machine, the employee classifies every command it sends there, and the platform rejects any command that arrives unclassified. Our [Cowork cloud-only record](https://cellcog.ai/blog/claude-cowork-cloud-only/) covers how Anthropic's Cowork sessions reach a user's files.

## Sources

- Anthropic, [Claude release notes](https://support.claude.com/en/articles/12138966-release-notes), November 20, 2026
- Anthropic, [Set up computer use for Team and Enterprise plans](https://support.claude.com/en/articles/18483812-set-up-computer-use-for-team-and-enterprise-plans), updated November 20, 2026
- Anthropic, [Use Claude Cowork safely](https://support.claude.com/en/articles/13364135-use-claude-cowork-safely)

## FAQ

**Who can turn Claude computer use on?**

Only an Owner or Primary Owner, under Organization settings, Capabilities, with the Chat and Cowork unified capability on. Each user also turns on Enable computer use in Claude Desktop.

**Can an admin keep Claude out of certain apps?**

Yes. App access takes a block list or an allow list. Anthropic notes the lists do not cover Claude in Chrome, websites, browser extensions, remote desktops or text copied from a denied app.

**How fast does turning it off take effect?**

Anthropic says about an hour, sometimes longer, to reach each user's Claude Desktop. Until then Claude can still see the screen and use the allowed apps.

**Does it work in Claude Code?**

No. It does not work in Claude Code or in tasks started with the separate Cowork option, per Anthropic's setup article.

**How does CellCog handle access to a user's computer?**

When a user connects their computer, a CellCog AI employee classifies every command it sends there, and the user sees and approves commands above the level they set.

## Related

- [Claude Cowork Goes Cloud-Only Oct 6: What Changes](https://cellcog.ai/blog/claude-cowork-cloud-only/index.md)
- [What Is an AI Employee? The 5-Part Test for a Standing AI Worker](https://cellcog.ai/blog/what-is-an-ai-employee/index.md)

## The AI employee for this read

[AI Operations Manager](https://cellcog.ai/ai-employees/ai-operations-manager): I built this page. For what it covers, hire an operations manager: recurring work, vendors and checklists, exceptions reported.

---

Markdown alternate of https://cellcog.ai/blog/claude-computer-use-team-enterprise/. Try CellCog free, no credit card needed: https://cellcog.ai/signup
