What happened
Anthropic’s announcement says Claude Meetings can join a video call when someone invites it or has auto-join enabled. Support is currently limited to Google Meet; the company says support for other conferencing platforms is coming. Once present, Claude can respond to questions asked aloud or in the meeting chat. Its default response is written, and it can use a participant’s connected tools with that person’s permission. Anthropic describes a workflow in which Claude drafts a plan from the conversation, retrieves prior event costs after asking for approval, compares them with the current plan, and turns a discussion of timelines into a whiteboard that attendees can edit while the call continues.
Claude also creates notes and action items. After a call, the notes, transcript, plan, whiteboard, and other artifacts appear on a meeting-specific page in claude.ai. People from the organization who were invited can see and edit the work; it can also be shared with others in the organization. Participants can later ask Claude about meetings they attended or add it to another meeting from their calendar. Anthropic says it is working with Gong, Attention, Granola, Fireflies.ai, and Otter on sending meeting transcripts to those services; the announcement describes this as upcoming, not an already available capability.
The announcement describes controls for calls with outside guests. By default, Claude stays out if an outside guest is on the invite and leaves if one joins during the call. An administrator can permit Claude to join calls with guests. In that case, Claude posts a notice in the call chat but does not reply there; guest contributions are included in the transcript and notes. Claude posts a notice when it joins any call. It saves a transcript, summary, and screen descriptions, but not audio. When someone asks a question aloud, Claude currently asks each time to confirm who is speaking. A yes to connected-tool access applies going forward after the first permission request.
The beta is rolling out to Team and Enterprise plans. On Team, inviting Claude is enabled by default and an administrator can turn it off. On Enterprise, an administrator enables it for all or part of the organization. Anthropic says some organizations cannot use the beta, including those with zero data retention (ZDR), HIPAA-ready configuration, or customer-managed encryption keys (CMEK). During beta, notes, transcripts, and answers are free on both plans unless an answer uses connected tools, which counts toward plan usage. Documents, whiteboards, and other meeting artifacts are free through February 10, 2027.
What the archive adds
New Runtime’s earlier coverage described Anthropic making unified Claude Chat and Cowork generally available to Enterprise customers, including the October 24 note. Claude Meetings is a new event: it extends Claude’s work from a unified chat-and-work interface into the live meeting itself, rather than restating that Enterprise rollout.
Trend signal
This supports agent-ready software exposes capabilities: Claude can act through connected tools, but the announcement makes permission, participant visibility, and organization-level controls part of the product boundary. The pattern is relevant, though the announcement does not establish how reliably those controls work in practice or how broadly teams will adopt them.
Why now
The beta is available now to eligible Team and Enterprise organizations, and Anthropic has specified which features are free and for how long. That makes the current period useful for teams to assess whether a live meeting assistant fits their consent, retention, and tool-access policies before considering wider rollout.
Connections to other events
New Runtime’s earlier coverage of Claude Cowork and chat coming together for Enterprise described an interface that combines conversation and task-oriented work. Meetings carries a similar idea into a different setting: a live discussion can produce editable artifacts rather than ending at a chat transcript. Anthropic also names Gong, Attention, Granola, Fireflies.ai, and Otter as prospective transcript partners, but does not say these integrations are live.
Why it matters
For teams, Claude Meetings changes the timing of AI assistance: a model can help shape a plan while participants are still discussing it, rather than summarizing only after the call. That could reduce the handoff from conversation to document, but it also makes the meeting itself a place where permissions and information exposure matter. A transcript and screen descriptions can preserve sensitive discussion even when no audio file is retained.
The defaults are consequential. External guests keep Claude out unless an administrator changes the policy, while a permitted guest call includes what those guests say in the notes and transcript. Teams should therefore treat auto-join as a data-governance decision, not just a convenience setting. Tool calls also have a cost boundary: ordinary answers are free in beta, but answers that use connected tools count toward plan usage. Eligibility excludes organizations relying on several security configurations, which may prevent some enterprises from testing the feature at all.
Limits and uncertainty
Anthropic describes a beta, not a generally available service, and limits current support to Google Meet. It does not report adoption, transcription accuracy, meeting-time savings, or the quality of generated action items and artifacts. The announcement does not establish how the product performs in noisy calls, with overlapping speakers, or across languages. Compatibility with other conferencing platforms and the named transcript partners remains prospective. Organizations should also verify whether their configuration qualifies before planning a trial; ZDR, HIPAA-ready configurations, and CMEK are specifically excluded.
Skill to build now
Before enabling an AI meeting participant, write a short meeting policy that answers three concrete questions: who can enable auto-join, how attendees are notified and consent is recorded, and who can access or share the transcript and artifacts afterward. Test that policy in a non-sensitive Google Meet call, including an outside guest, before allowing the assistant into routine meetings.
Reader takeaway
Claude Meetings makes the meeting a live workspace for questions, documents, and follow-up—not just a source for a later summary. The useful first step is to test whether its permission and transcript behavior fits your organization’s rules, especially for outside guests and connected tools.