- Updated: March 24, 2026
- 2 min read
Anthropic Unveils Claude Code and Cowork: AI Agents Now Control macOS in Research Preview
Anthropic has rolled out a new research preview of Claude Code and Claude Cowork, two AI‑driven tools that let large‑language models write, run, and debug code while directly controlling a macOS computer. The announcement, first reported by The Verge, highlights a significant step toward more autonomous AI agents that can interact with graphical user interfaces, move files, and even launch applications without human intervention.
Both Claude Code and Claude Cowork are built on Anthropic’s Claude 3 model. Claude Code focuses on software development tasks—generating code snippets, fixing bugs, and running scripts—while Claude Cowork extends those capabilities to broader computer‑control actions such as opening Safari, navigating Finder, and managing system settings. The tools are currently limited to macOS, but Anthropic hints at future support for Windows and Linux.
The features are offered as a research preview, meaning they are still experimental and subject to change. Users must join a waitlist and agree to usage policies that restrict potentially risky actions, like executing arbitrary commands or accessing sensitive files. Anthropic emphasizes safety, noting that the models are designed to ask for clarification before performing high‑impact operations.
For developers and businesses interested in AI‑augmented workflows, these capabilities open new possibilities: automated testing, rapid prototyping, and even AI‑assisted remote work. The tools also align with the growing trend of “AI agents” that can act autonomously across software ecosystems, a concept explored in depth on our AI Agents hub.
Anthropic’s move adds fresh competition to the AI‑agent space, where companies like OpenAI, Google, and Microsoft are also experimenting with code‑writing assistants and UI‑level automation. While the current macOS‑only limitation narrows the immediate audience, the research preview will provide valuable data on how users interact with AI‑driven computer control.
Stay tuned for updates on Claude Code and Claude Cowork as Anthropic expands functionality and opens the tools to a broader developer community. For more Anthropic‑related news, visit our dedicated page at Anthropic News.
Andrii Bidochko
CTO UBOS
Andrii Bidochko is an AI entrepreneur and researcher focused on AI agents, reinforcement learning, and autonomous systems. He writes about the technologies shaping the future of machine intelligence, from frontier models and agent architectures to real-world AI applications.