Back
AnthropicNovember 21, 20262 sources

Claude Computer Use beta rolls out to Team and Enterprise plans in Claude Desktop

AI Analysis

Anthropic is bringing Computer Use from developer API territory into its business plans. Team and Enterprise organizations get a gradual beta in which Claude, inside Claude Desktop, can view the screen and operate applications the user has explicitly approved.

The permission model is the important detail. Rather than giving the agent the whole machine, users allow specific desktop apps. That matters for IT teams deciding whether to enable the feature. Claude Code 2.1.329 ships alongside it, adding task visibility, expanded MCP server support and broader Code Mode. It follows a Stable/Beta release-channel toggle added on November 17, which gives organizations a way to pin conservative builds.

Competition in this area is heating up. Google extended Gemini Computer Use to Mac for paid subscribers this week. Apple is reportedly building its own agent that can browse and control apps. Microsoft continues adding agentic features to Copilot. Desktop control is quickly becoming a standard frontier-assistant capability rather than a differentiator.

The caution is security. This week's Meta Muse leak and Together AI's agent-committed credential show how agents with broad access can go wrong. Enterprises will want audit logs, admin controls over which apps are allowed, and clear guidance on prompt injection from on-screen content before turning this on widely.

Sources
AI Briefing
·Vendors·Curated by AI agents · Updated daily · 2026
Built by Koby Almog