Anthropic takes its full agent stack out of beta in one move

Anthropic has moved several pieces of its agent toolkit out of beta on the same day: computer use, a browser-use tool, the Skills API, and the Files API. Taken together, this is about making AI agents more practical for work that happens inside real business software.

The distinction matters. Computer use lets an agent interact with software through the interface, much as a person would. The browser-use tool is separate. It works through a browser’s structured accessibility tree, rather than relying on visual screen interaction alone. That gives developers another way to build agents that can navigate web-based systems. Skills and Files add the supporting pieces for giving agents reusable capabilities and access to the documents they need to work with.

Moving the full stack out of beta is a commercial signal. Anthropic is saying these tools are ready to be used in production, not just tested in controlled demos. For owners and operators, the opportunity is not an AI that writes another draft email. It is an AI that can help carry out defined steps across the systems your team already uses, with the right guardrails around it.

The hard part now is durability. An agent has to handle messy files, changing interfaces, unclear instructions, and the exceptions that show up in everyday operations. Anthropic has set a bar that OpenAI and Google will need to match. This is the kind of thing we build into an AI command centre, connecting useful AI actions to the actual workflows that run the business.