Anthropic has expanded Claude’s Gmail integration so the assistant can take direct action inside a user’s inbox, according to Digital Trends. The new capability lets Claude send, reply to, and forward Gmail messages on a user’s behalf, including without stopping for approval before every outgoing message when that setting is enabled. The update builds on Claude’s existing Google Workspace connector, which Digital Trends says already supported Gmail, Google Calendar, and Google Drive. The missing piece was outbound email: Claude could help work with messages, but sending was the more consequential action. With the latest change, users can ask Claude to respond to an email thread, have it prepare the reply, and send it. The control layer is the important part. Digital Trends reports that Anthropic says approval remains the default. Users can decide whether Claude must ask for confirmation before sending emails, while Team and Enterprise administrators can determine whether members are allowed to let those actions happen without repeated approval. The feature is limited to paid Claude plans, according to the report. It also includes forwarding emails, which broadens the integration beyond drafting or summarizing and moves Claude closer to acting as an inbox operator for routine communication. That shift changes the risk profile. Generating a suggested reply is reversible until the user presses send; authorizing an AI assistant to send directly means an error can become an external communication. Digital Trends frames Anthropic’s approval system as the safety valve, but the report also notes that disabling repeated approval puts more responsibility on users and workplace administrators. For operators, the practical use case is repetitive email: routine replies, follow-ups, and forwarding messages. The value is not that Claude can write email text, which AI assistants have been able to do for some time, but that it can complete the workflow inside Gmail when permitted. The broader signal is that Claude’s Workspace integration is moving from retrieval and drafting toward action. Based on the provided report, Anthropic is keeping the highest-friction default in place, while giving paying users and enterprise admins a way to trade manual review for speed in the inbox. Who benefits: Paid Claude users who handle repetitive Gmail workflows could save time on replies, follow-ups, and forwarding. Team and Enterprise administrators also get a control point for deciding whether that autonomy is allowed inside their organizations. Who's exposed: Users and organizations that disable repeated approval take on the risk of Claude misunderstanding context and sending an incorrect message. The exposure is highest where email content is sensitive, external-facing, or subject to workplace review norms.