Anthropic brings Claude into live meetings and shared work
Claude Meetings joins Google Meet calls as a visible participant, turning spoken discussion into transcripts, action items and shared artifacts.
From meeting recorder to participant
Anthropic has introduced Claude Meetings, a feature that places Claude directly inside a team’s video call. The launch begins with Google Meet, with support for additional conferencing platforms planned later.
Unlike a passive transcription service, Claude appears as a visible participant. People can ask it questions verbally or through chat while the meeting is in progress, and can ask it to begin creating a document, whiteboard or other artifact during the discussion.
After the call, Claude provides a meeting page containing notes, action items, the transcript and anything it created. That structure turns the meeting record into a working space rather than a file that is merely archived.
A stronger test for agent usefulness
The product is significant because it moves Claude into a high-frequency coordination setting where context changes quickly and mistakes are socially visible. A meeting agent must identify decisions, separate tentative ideas from commitments and preserve ownership across several speakers. It also has to remain useful without interrupting the conversation or producing a second stream of work that nobody maintains.
Anthropic’s choice to make Claude a visible participant matters for trust. Colleagues can see when it is present, question its output and correct artifacts while the discussion is still happening. That is a different product posture from silently recording a call and delivering an opaque summary afterward.
Why it matters
The launch extends Claude from an individual assistant into a collaborative interface for teams. If the generated artifacts remain accurate and editable, meetings could become an entry point for delegated work rather than the end of a conversation. The open risks are familiar but amplified: consent, confidential discussion, speaker attribution and the cost of correcting a polished but wrong action list.
Uncle Cat take: The visible-participant design is the important constraint here; Anthropic is making correction part of the workflow instead of hiding the agent behind a transcript.