Claude in Chrome is generally available
Claude in Chrome is now generally available on every paid Claude plan. Claude can take actions autonomously in the browser instead of needing approval for each one. A safety classifier validates every action, and probes screen web content for prompt injections first.
Why it mattered The browser agent graduates from pilot to product: autonomous browser actions on paid plans, backed by evaluated prompt-injection defenses.
Anthropic made Claude in Chrome generally available on all paid Claude plans. The extension lets Claude view the current page and act on it: reading and typing text, clicking links, navigating, and filling out forms with the user’s existing logins.
The headline change from the pilot is autonomy. Claude can now take actions without per-step approval, using the same automatic-approval mechanism as auto mode in Claude Code. A safety classifier checks each action against the original request and blocks mismatches. Anthropic reports its prompt-injection evaluations found no successful attacks against Sonnet 5, Opus 5, or Mythos 5 with probes plus classifiers on.