Claude Meetings
Canonical version: Claude Meetings.
Claude Meetings lets Claude join your video calls as a visible participant. Announced by Anthropic on 2026-11-12, in beta for Team and Enterprise plans (rolling out gradually), and Google Meet only for now (other platforms are "coming soon", with no names or dates).
Why is it interesting? AI note-takers already exist (Granola, Otter, Fireflies.ai, Zoom's built-in notes...). They all do the same thing: listen, then hand you a summary AFTER the call. Claude Meetings goes one step further: it does work DURING the meeting. Ask it a question, and it can look up a number in a connected tool and post it in the chat while you're still talking. Ask for a doc or a whiteboard, and it builds it live, and everyone on the call can edit it.
How it works
- You invite Claude to a meeting (or turn on auto-join). You can add it to upcoming meetings from your calendar in claude.ai
- It shows up as a participant and posts a notice in the call's chat, so nobody is recorded without knowing it
- You ask questions out loud or in the chat. Claude answers in writing by default
- It can use your connected tools, with permission. Each person authorizes their own tools; Claude asks the first time, and the approval carries forward. For spoken requests, it confirms who's speaking every time (smart, since a voice is easy to fake)
- It produces notes, action items, a transcript, a summary and a recap (which you can share before the call even ends)
- Each meeting gets a page in claude.ai with the notes, the transcript and everything created during the call. Invitees from your organization can view and edit it
- Later, you can ask Claude about past meetings you attended
What gets saved: the transcript, the summary and descriptions of what was on screen. NOT the audio.
Outside guests and admin controls
By default, Claude stays out of any meeting with an outside guest, and leaves if one joins. Admins can allow outside guests. In that case, Claude still posts its notice but doesn't reply in the chat, and what guests say still ends up in the notes and the transcript.
On Team plans, inviting Claude is on by default (admins can turn it off). On Enterprise, admins enable it for all or part of the organization. Organizations with zero data retention (ZDR), HIPAA-ready setups or customer-managed encryption keys (CMEK) can't use it at all.
Pricing (beta)
- Notes and transcripts: free on Team and Enterprise
- In-call answers: free, unless they use connected tools (those count toward your plan's usage)
- Docs, whiteboards and other artifacts: free until 2027-02-10
Nothing is said about pricing after the beta.
Funny detail: Anthropic lists Granola, Otter, Fireflies.ai, Gong, Attention and Plaud as partners, and you'll "soon" be able to send transcripts to them. So the note-taker apps it competes with are also its integration partners.
Reception on Hacker News
The Hacker News thread (100 points, 76 comments) talked more about consent and power than about features:
- Consent: there's a consent click-through when you join; refuse and you don't join the meeting. Some saw that as fine, others as a choice that isn't really one. One commenter's take: "This ship has long since sailed"
- Security: "It's going to be very tempting to give tool access to these agents", and the security model behind that is hard. An agent that listens to anyone in a call AND has tool access is a textbook case of Prompt injection risk (see Lethal Trifecta for AI Agents)
- Platform risk: "They are really coming for Microsoft and Goog." Several people asked why partners would keep building on Anthropic if it absorbs every profitable integration, and whether feeding your roadmap to a model vendor is wise
- Meetings themselves: some argued most meetings are overhead that AI should kill. Others replied that "some human needs to understand and decide at least something"
My take
I like the direction: an assistant that acts during the meeting is much more useful than yet another summary that nobody reads. It's the same "Claude as a teammate" pattern as Claude Tag in Slack, applied to calls, and it complements Claude Duet (which works on your behalf between meetings).
But tooling doesn't fix bad meetings. Meetings need to be designed: a clear goal, the right people, a decision at the end. If the meeting has none of that, a perfect transcript of it is still worthless.
References
Related
About Sébastien
Ready to get to the next level?
Found this valuable? Share it with someone who needs it.