Anthropic has pushed Claude from chatbot to actual operator, giving the model real browser control: it can open pages, click, type, scroll, and complete multi-step web tasks the way a human would. The move follows Anthropic's earlier "use your computer like a human" rollout and its Claude Code push toward non-programmers, and it lands right as Indian teams already lean on Claude for everything from ITR filing to research.
The shift matters because most Indian business workflows still live inside a browser: GST portals, bank dashboards, vendor logins, insurance claims, and lead follow-ups across five different SaaS tools. An agent that can actually drive those pages removes the copy-paste tax that has stalled every RPA project of the last decade. The catch is the same one that sank those projects: permissions, credentials, audit trails, and who owns the mistake when the agent clicks the wrong button on a live account.
- Core Update: Claude can now control a real browser end to end, navigating, clicking, typing, and completing web tasks instead of just describing them.
- Key Metrics / Specs: Built on the same operator stack Anthropic introduced with "Claude can use your computer like a human" and extended through Claude Code. No public per-action pricing confirmed yet; usage flows through existing Claude plans and API tiers.
- Access & Availability: Rolling out to Claude users and API customers now, with enterprise controls and permission guardrails as the deciding factor for company-wide deployment.
- India Context: Strong local pull already: professionals compared Claude-assisted ITR filing to "a CA sitting next to me," and banks are under fresh pressure on cyber resilience after the Claude Mythos conversation.