Claude Cowork now ships with its own browser for AI‑driven web work

Anthropic has also upgraded Claude Cowork, its enterprise‑oriented AI workspace, with a built‑in browser that lets Claude act directly on the web instead of just generating text. Announced on August 26, the new feature turns Cowork into a place where AI agents can open sites, navigate pages, fill in forms and collect data as part of multi‑step workflows. In practice, it’s Anthropic’s take on the “computer‑use” or browser‑agent trend, but packaged for corporate environments.

The browser can blend internal knowledge sources with public web content. A user might ask, “Compare pricing and feature tiers for competitors A and B and put the results into a spreadsheet,” and Claude will browse relevant sites, extract tables, and assemble a structured document without the user manually clicking through pages. Typical scenarios include market and competitive research, sourcing candidates from job boards, or scanning regulatory sites for updates linked back to in‑house policy docs.

Anthropic stresses that activity inside the browser is governed by enterprise security controls: organizations can define which domains are accessible, how logs are stored and which data can be retained for model improvement. The browser is rolling out to existing Claude Cowork customers at no additional cost, giving them a way to test agentic web automation without deploying separate RPA or headless‑browser stacks.