Newshot
Account

Claude Joins the Meeting: Documents and Tasks Take Shape Before the Discussion Ends

AI summary

Claude now joins Google Meet as a visible participant, and documents, whiteboards, notes, and action items take shape while the discussion is still ongoing. By default it answers only in text and asks a participant for permission before using tools connected to that person; it stores text transcripts, summaries, and screen descriptions rather than audio, and when someone outside the organization is invited it does not join by default, exiting if an external person joins mid-meeting. Which meetings it can join, what it can see, and who approves tool use are still governed by plan, permissions, and guest rules, and admins can relax the external-participant restriction, in which case it only posts notifications in chat.

Translation(model translation, not the source's own text)

Claude Joins the Meeting: Before the Discussion Ends, Docs and Tasks Take Shape

Claude Meetings is opening in Beta to Team and Enterprise teams, joining Google Meet as a visible participant that listens along and generates documents, whiteboards, notes, and action items; by default it answers in writing only, and is bound by permission and recording rules when calling connected tools or handling external guests.

Cover image from Claude: Blog (official article image)
Cover image from Claude: Blog (official article image)
  • Claude follows the conversation as a participant and drafts outputs in sync
  • Invited teammates can co-edit the plans formed in discussion during the meeting
  • Calling a personally connected tool requires that person's permission first
  • Every meeting saves a transcript, summary, and screen descriptions

This article is based on Anthropic's official product blog and is vendor release information; features, plans, platform support, and pricing are subject to the official current descriptions.

Claude now enters Google Meet as a visible participant—before the discussion ends, documents, whiteboards, notes, and action items have already taken shape. By default it answers in text only, and asks a participant for permission before calling a tool that person has connected; when a meeting invites people from outside the organization, it does not join by default. Which meetings it can enter, what it can see, and who approves tool calls are still determined by plan, permissions, and guest rules.

What exactly does Claude do while sitting in the meeting?

Claude Meetings lets Claude enter a video meeting as a visible participant. It does not speak and by default answers only in text, but the team can call on it at any time during the discussion.

It joins as a visible participant (a member directly visible in the meeting interface) and posts a notice in the meeting chat. Team members can ask out loud or type in the chat. By default, it answers in text and does not speak.

It can use the tools you have connected, provided it gets permission. It asks the first time, and once granted, the approval continues to apply. Verbal questions currently have one hassle: it must confirm who is speaking each time. Answers posted in the chat are visible to everyone in the meeting.

It saves the transcript, summary, and screen descriptions—not audio. Under default settings, if the invitation includes anyone from outside the organization, it will not join the meeting; if an external person joins midway, it will leave. Administrators can lift this restriction, in which case it only posts a notice in the chat and does not reply in the chat.

Before the meeting, confirm that everyone present accepts being transcribed.

Before the meeting ends, the plan is already editable

Claude Meetings organizes the event plan the team has discussed into working material that can continue to be edited during the discussion.

The official example is quite specific. The team is planning a customer event, and Claude is on the call too. Someone says directly: "Claude, start writing a document for this plan." Claude writes up what the team has said so far, and invited teammates can edit the plan while the discussion continues.

This plan is not a locked final product once handed to Claude. When discussing the budget, someone asks Claude how much last year's event cost; it gets consent first, then looks it up, and posts the result in the meeting chat to compare with the current plan. When the conversation turns to scheduling, someone asks it to put it on the whiteboard, and the team schedules together next to the document.

After the meeting ends, the notes, action items, plan, and whiteboard remain on the meeting page in claude.ai. Later you can return to meetings you attended and follow up with Claude, and you can add it to the next meeting from the calendar.

Claude: Blog (web) official image 1
Claude: Blog (web) official image 1

What gives it the right to read someone else's tool data?

Before Claude calls a tool that a participant has connected during a meeting, it does not treat "the tool is connected" as "it can be read freely." It first asks the person who owns the tool. Only after approval does it look up the information and post the answer in the meeting chat.

Permission is granted per person, not uniformly per meeting. Claude asks the first time it uses someone's connected tool; once that person agrees, the approval can continue to apply, with no need to reconfirm each time. It uses each person's own connected tools, so it must obtain that person's permission.

Verbal naming adds another identity check. Currently, every time Claude is called on verbally, it reconfirms who the speaker is. The reason is straightforward: multiple voices are connected in a meeting, and it cannot judge from a single "Claude" whose request it should be attributed to. The tool's data and permissions are likewise bound to the person who granted approval.

After getting permission, Claude looks up the needed information and posts the result in the meeting chat. The official example: when the team discusses the budget, someone asks about last year's event spending; Claude first requests permission to look it up, then posts the result in the chat for participants to view and compare together.

The chat area is not a private window visible only to the questioner. Answers Claude posts are visible to all participants in the meeting, and content it is permitted to read may also be displayed in front of all participants. Whether tools can be connected is not the point—the person granting approval and the answer being public are the bottom line of this process.

Claude: Blog (web) official image 2
Claude: Blog (web) official image 2

After the meeting, the record can still be edited

Claude creates a separate page for each meeting in claude.ai, gathering the written material left by the discussion into it. What it saves is content that can continue to be processed, not meeting audio.

The page includes meeting notes, the transcript, and content Claude created during the meeting. After the meeting, team members can return directly to the corresponding page to view and edit, without piecing material together segment by segment from the chat log.

First, the boundaries. Officially, before enabling automatic joining, team members need to confirm that all participants accept being transcribed—this feature processes meeting content from the very start. This is not a footnote; it determines what can go on the page and who can see it.

The page uses a permission scope for collaboration within the organization: invited members can view the page and edit content Claude created; to give it to other colleagues, you can share it directly from the meeting page with others in the organization. Sharing is limited to within the organization and should not be understood as the meeting page being open to external guests.

Three types of content are saved: the transcript, the summary, and screen descriptions. The transcript restores the discussion content, the summary provides a condensed meeting thread, and screen descriptions record information Claude can recognize on the shared screen. Audio itself is not within the saved scope—the page leaves textual material of the discussion and screen, and cannot play back recordings.

These records can also carry work forward: notes and content such as plans Claude created remain editable, not a read-only meeting archive.

If external guests join, can Claude still attend?

By default, no. As long as the meeting invites anyone from outside the organization, Claude stays out of the meeting; if an external person joins midway, Claude also leaves and will not continue listening in or processing the discussion.

Administrators can change this setting to allow Claude to attend meetings with external guests. This allowance is an explicit administrator setting, not a participant's temporary verbal consent; under the default state, external guests still trigger the exit.

Once permitted, Claude appears in the meeting as a participant and posts a notice in the meeting chat so external guests know the AI is present. It does not reply to questions in the chat, but what external guests say still enters the notes and transcript Claude generates, and the team can see it later when viewing the meeting page.

Who can use Claude Meetings? Four conditions to check first

Claude Meetings is opening in Beta to Team and Enterprise plans, but "can join meetings" does not mean every team and every organization can use it directly. Before enabling, check the platform, administrator switches, data configuration, and usage rules.

The Team plan enables inviting Claude by default, and administrators can turn it off. For the Enterprise plan, administrators decide whether to enable it for all or part of the organization. Team leads should check whether administrators have changed the default settings; Enterprise administrators should also confirm the scope of inclusion, so members do not discover the feature is not enabled only after initiating an invitation.

Currently only Google Meet is supported; other video meeting platforms are not yet open. Teams preparing to use Claude on Microsoft Teams, Zoom, and other platforms cannot temporarily use the same meeting process. The official statement says other platforms are coming soon, but no specific date is given.

Beyond the platform, there is a harder restriction: one category of organizations cannot use it for now, including organizations with zero data retention enabled (not allowing the platform to retain usage data), HIPAA-ready configurations, or customer-managed encryption keys (where the customer controls the data encryption keys). HIPAA-ready refers to compliance configurations prepared for scenarios related to the U.S. Health Insurance Portability and Accountability Act. Administrators should first confirm whether the organization hits these restrictions and should not assume availability merely because it subscribes to Enterprise.

Billing during Beta also has boundaries. Meeting notes, transcripts, and in-meeting answers are free for Team and Enterprise; if an answer calls an external tool connected by a team member, it counts toward plan usage. Documents, whiteboards, and other outputs Claude creates during meetings are free until February 10, 2027.

Source: Claude's official product blog "Introducing Claude Meetings" (November 12, 2026); the features, restrictions, and free period in this article follow the Beta description at release, and feature and plan status may be adjusted afterward.

Glossary · Plain-language notes in this piece

The full text of deep dives, research analyses, and tool tutorials, plus all podcasts, are in the member area.

The translation is machine-generated and individual terms may be off. Nothing is rewritten or cut; the original is authoritative.

Published by 听说了吗
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗