Anthropic Adds Write Actions to Claude's Google Workspace Connectors

On August 18, 2026, Anthropic expanded its Google Workspace connectors to allow Claude to send, reply to, and forward emails in Gmail. [3] The update also enables the AI to share, move, and trash files within Google Drive. [3] These new write actions feature human-in-the-loop approval enabled by default. [3] For users on Team and Enterprise plans, organization owners control whether members are permitted to skip these per-action confirmations. [3]

Alongside the connector update, Claude Cowork was rolled out to all paid accounts across mobile and web platforms. [2][3] The Gmail connector page indicates compatibility with Claude Code and the Claude API, in addition to standard chat interfaces. [3] This allows write-capable email access to inherit directly into coding sessions. [3]

This update shifts Claude from a passive reader of workspace data to an active participant, pushing it into closer competition with Microsoft's agentic Copilot capabilities in Microsoft 365. By mandating human-in-the-loop approvals by default, Anthropic acknowledges the severe risks of automated email and file deletion, though enterprise admins can choose to bypass these safeguards. The integration of write permissions into Claude Code signals a push toward fully autonomous coding agents that can email reports or push files independently. What the sources don't address: whether Anthropic will introduce granular permissions that allow admins to enable email drafting while keeping file trashing permanently disabled.