Skip to content

Anthropic

Claude gets write access to Microsoft 365 on Dec 9

3 min read

Written by an AI model from the linked sources and checked against them by a second model. How we work

Image: support.claude.com

Starting December 9, 2026, Claude will be able to send email, change calendars, edit files and post in Teams for many people who sign in with a Microsoft 365 work or school account. Until now it could only search and read their data unless an admin had turned on its write tools. On or after that date, write tools that are still blocked by default will be switched on, according to Anthropic’s Claude support page. They will be set to “Needs approval”, and admins who have already blocked them keep them off.

The connector launched in February as a read-only tool for Outlook, OneDrive, SharePoint and Teams. In April, Anthropic opened it to every plan, including the free one, Vucense reported. Since August, Claude has also been able to send Gmail replies and manage Drive files.

What changes on December 9

With write tools on, Claude can draft and send emails, create and manage calendar events, create and update files, and send Teams messages, including chat messages, channel posts and replies, and new chats. It acts only within the user’s existing Microsoft 365 permissions. Even with write tools on, it can’t attach files to emails or change Teams settings or permissions, and each user has a limit on write actions.

Emails Claude sends carry a header that marks them as agent-initiated. Teams messages, file changes and calendar changes aren’t tagged for now, so the people receiving them can’t tell Claude made them.

Personal Microsoft accounts such as Outlook.com or Hotmail, which can be connected on Free, Pro, Max and Ultra plans, already have write tools turned on unless the user switches them off under Tool permissions.

When Claude asks first

By default, in chat and Cowork, Claude asks you to confirm every time it sends or forwards an email, sends a Teams message, creates, changes, deletes or responds to a calendar event, or sets an automatic reply.

Things change in “Automatically approve” mode. There, these actions can run without asking unless an automated check flags them as risky or your organization requires approval, and Anthropic warns that the check can miss risky actions. If an admin sets one of these tools to Always allow, Claude may use it without asking anyone. That applies in “Automatically approve” mode and in some scheduled tasks, such as a cloud scheduled task set to “Automatically approve” or a Claude Code routine.

On Team and Enterprise plans, Claude keeps asking unless an admin set the tool to Always allow on or after November 9, 2026, or the organization first added the connector after that date. In those cases users can choose Always allow themselves if their admin permits it, and an Always allow choice a user made in the past can take effect right away. These exceptions don’t apply in organizations that have enabled HIPAA.

What admins can do

The tools stay off only where an admin has set them, or the whole connector, to Blocked. Admins who want to keep Claude read-only should do that before December 9. Users who want to approve every action themselves can switch to Manual mode or set a tool’s permission to “Always ask” where that option is available.

No comments yet. Be the first.

Join the conversation

Sign in to leave a comment. It appears after moderation.