An AI Colleague Joins the Call: Anthropic Launches Claude Meetings in Beta
On Nov 12, 2026, Anthropic launched Claude Meetings in beta: Claude joins Google Meet calls as a visible participant, writing docs, whiteboards and notes while the meeting runs. Team and Enterprise only, with outside guests kept out by default.

What happened: Claude has officially taken a seat in your video calls
On November 12, 2026, Anthropic announced that Claude Meetings is starting to roll out in beta: Claude can join your video calls as a visible participant, follow the conversation, and start working while the meeting is still going. It answers questions raised in the meeting, drafts documents, draws timelines on a whiteboard, and writes the meeting notes and action items. After the call, the notes, action items, documents, and whiteboards all appear on a dedicated page for that meeting in claude.ai.
This is not an "automatic notes after the meeting" tool. There are already plenty of those: Otter, Fireflies, and Granola all record first and organize later. The difference with Claude Meetings is in the words "during the meeting": it starts working while the discussion is still happening. Anthropic's blog gives a concrete example. A team is planning a customer event on a call, and someone says, "Claude, start a doc for this plan." Claude writes up what the team has said so far, and the people on the invite can edit the plan while the conversation continues. When the budget comes up, someone asks what last year's event cost; Claude asks for permission to look it up, posts the figure in the meeting chat, and compares it with the current plan. When timelines come up, someone asks Claude to put it on a whiteboard, and the team works on the schedule alongside the doc.
The rollout scope needs to be stated clearly first, so you don't go looking in vain. First, it is only for Team and Enterprise plans — the personal Free, Pro, Max, and Ultra plans don't get it — and it is "starting to roll out," so not every eligible account will have it on day one. Second, it only supports Google Meet for now; other video conferencing platforms are "coming soon," with no date given. Third, some organization types are excluded from the beta: those with zero data retention (ZDR), HIPAA-ready configurations, or customer-managed encryption keys (CMEK) can't use it yet.
What it actually does in a meeting: four things, all happening live
The first is answering questions. People in the meeting can ask Claude out loud, or in the meeting chat. By default, Claude answers in writing; if it needs to use one of your connected tools — for example, to look up what last year's event cost — it asks for your permission first. The first time it uses a particular person's tool it asks once, and a yes applies going forward; but when someone asks out loud, Claude confirms who is speaking each time. It is a deliberately restrained design: it knows it is acting for different people, and permissions must not leak across them.
The second is building docs and whiteboards while people talk. This is where Claude Meetings pulls away from traditional meeting-notes tools. Other tools work on the pattern "you finish talking, then I organize"; Claude works on "you talk, I write." The docs and whiteboards are live, and participants can keep editing them on the spot. After the call, these artifacts sit on the meeting's page in claude.ai together with the notes and action items, and later you can go back and ask Claude about a meeting you attended, or add Claude to the next one from your calendar in claude.ai.
The third is writing notes and action items so nobody on the call has to. Meetings used to need someone to "take notes on the side," and the note-taker was effectively half absent. With Claude taking that job, everyone in the meeting can focus on the discussion itself. Anthropic also says it is working with partners like Gong, Attention, Granola, Fireflies.ai, Otter, and Plaud, so you will soon be able to send a meeting's transcript to the tools your team already uses.
The fourth is traceability after the meeting. Every meeting Claude attends gets its own page in claude.ai, holding the notes, the transcript, and anything it built. People from your organization on the invite can see and edit what Claude made, and the page can be shared with others in your organization. This design turns meetings from "over and scattered" into "a reusable record," which is especially useful for projects that need to sync across teams.
Privacy by design: "visible" and "leave unless approved" are the defaults
Getting an AI into video calls faces its biggest obstacle not in features but in trust: will it secretly record? Will it send what I said to someone else? Anthropic spends a whole section of the announcement on security and privacy, clearly aware that this section decides whether the product succeeds or fails.
First, it is visible. When Claude joins a meeting, it appears as a participant and posts a notice in the meeting chat, so everyone can see it is there. It is not a hidden recording bot. Second, it keeps text, not audio. Claude saves a transcript, a summary, and screen descriptions — not audio. The announcement also states plainly: before you add Claude or turn on auto-join, make sure everyone on your calls is OK with being transcribed.
The most important rule concerns outside guests. By default, Claude stays out of calls when someone from outside your organization is invited, and leaves if someone from outside joins partway through. An admin can allow Claude on calls with outside guests, but even then there are constraints: on those calls, Claude posts its notice in the chat but doesn't reply there, and what guests say is included in the notes and transcript. The message is blunt: internal meetings can be opened up; external meetings are locked down by default.
The on/off defaults also differ by plan. On Team plans, inviting Claude to meetings is available by default, and an admin can turn it off; on Enterprise plans, an admin turns it on for all or part of the organization. During the beta, Claude's notes and transcripts are free on Team and Enterprise plans — unless an answer uses one of your connected tools, which counts toward your plan's usage. The docs, whiteboards, and other artifacts Claude makes in a call are free through February 10, 2027.
Why now: the last piece of Anthropic's office-suite puzzle
Claude Meetings did not appear out of nowhere. Two days earlier, on November 10, Anthropic released Claude Whiteboards, Claude Sheets, and Claude Data (formerly Claude Dashboards) as generally available, and in that announcement Meetings was described with a single line: "coming soon." On November 12, the last piece of the puzzle was filled in. Taken together, the four products make the intent unmistakable: Anthropic wants to take on Microsoft 365 Copilot and ChatGPT's workplace stack head-on in office scenarios.
There is another layer of context in the timing: Anthropic is in the middle of its IPO process, and in the same week it published an IPO price range and shipped a dense run of enterprise products. For a company about to go public, enterprise subscription revenue is the story the capital markets most want to hear — and meetings are the highest-frequency, hardest-to-replace part of enterprise software. Claude Meetings is not aimed at the small feature of "helping you write notes"; it is aimed at "wherever your team meets every day, Claude is there."
Seen competitively, Claude Meetings faces two groups of rivals. One is meeting-notes tools like Otter, Fireflies, and Granola: their advantage is neutrality — not tied to any one model vendor, able to record any meeting — and their disadvantage is that they only record, they don't do work. The other is AI embedded in office suites, like Microsoft Copilot and Google Gemini: their advantage is being closest to documents and calendars, and their disadvantage is that the model is not theirs to control. Anthropic is playing a third hand: the strongest model, able to do work live in the meeting, but starting with Google Meet only — betting that "ability to do work" matters more to teams than "platform coverage."
How it differs from "post-meeting notes" tools

The comparison chart above compresses the difference into five rows, and each is worth reading. The first row is "when does it work": traditional tools organize after the meeting; Claude starts during it. For people in meetings, that means by the time the meeting ends, the plan document already has a first draft, instead of "notes in three days." The second row is "what does it produce": traditional tools produce notes; Claude produces notes plus action items plus docs plus whiteboards — a set of working results that can be used immediately. The third row is "outside guests": traditional tools record anything; Claude stays out of meetings with outsiders by default — the price it pays for trust. The fourth row is "what is kept": traditional tools keep audio; Claude keeps only text and screen descriptions. The last row is "how it is priced": during the beta, notes, transcripts, and in-call answers are free on Team and Enterprise plans, and artifacts are free through February 10, 2027.
But the same chart exposes Claude Meetings' current weakness: it only supports Google Meet. If your team meets on Zoom, Teams, or Tencent Meeting, you can't use it today. Anthropic says other platforms are "coming soon," but gives no date and names no platform. Platform coverage is homework it has to do — otherwise the advantage of "working during the meeting" will be canceled out by "it can't join my meeting."
The reality for small teams and independent developers: first check whether you can even use it
If you are an independent developer or a team of three to five people, don't get excited just yet. Claude Meetings most likely has nothing to do with you today: it is only on Team and Enterprise plans, and the personal Free, Pro, Max, and Ultra plans don't get it. Third-party summaries of pricing put the Team plan at $25 per seat per month, or about $20 billed yearly — a real cost for a small team, and you should verify it against the official pricing page before buying.
Even with a Team plan, there are two more gates. The first is platform: only Google Meet is supported. If your team meets on Zoom or Tencent Meeting, Claude Meetings can't join today. The second is compliance: organizations with zero data retention, HIPAA-ready configurations, or customer-managed encryption keys are excluded — and those are exactly the healthcare and finance industries where meeting notes matter most. Anthropic's trade-off is clear: it would rather cover fewer scenarios at first than risk its reputation on privacy.
So what can a small team learn from this? At least three things can be copied directly. The first is the product idea of "output during the meeting": stop treating meeting notes as a post-meeting accessory, and try letting AI write conclusions into a document while the discussion is happening, with participants editing on the spot, so the meeting ends with a usable draft. The second is the permission design of "outside guests denied by default": for scenarios involving people outside the company, locked by default wins more trust than open by default. The third is the retention policy of "keep text, not audio": you don't need to store every piece of raw material — the less you keep, the more willing people are to use it.
A seven-step trial checklist: once you have access, don't roll it out to everyone at once
If your team happens to be on a Team or Enterprise plan and has received beta access, follow these seven steps rather than enabling it for everyone on day one. First, the admin should confirm the Meetings switch in Organization settings: Team has it on by default and can turn it off; Enterprise has it off by default and needs to turn it on — find out which side you are on. Second, pick one internal weekly meeting as a pilot; don't start with a meeting that has outside clients. Third, tell the group before the meeting: Claude will be attending, the meeting will be transcribed, and anyone who disagrees can say so in advance.
Fourth, have one person in the meeting be responsible for asking Claude to do things — "write a plan document," "put the timeline on the whiteboard," "look up last year's spend" — instead of everyone shouting at once. Fifth, after the meeting, open the meeting page in claude.ai and check the notes and action items for errors: names, numbers, and deadlines are the three things most likely to be wrong. Sixth, write down the errors, and keep a record for three meetings in a row; if the same kind of error keeps recurring, pause usage until the vendor fixes it. Seventh, after a full month, decide: count how many meetings produced artifacts that were actually used afterward, instead of sitting unread on a page. That number is the only standard that matters for renewal.
Judgment and boundaries: it is betting that "ability to do work" matters more than "platform coverage"
This article's judgment is that Claude Meetings' biggest selling point is not "it can join meetings" but "it can do work once inside." The competition in meeting-notes tools is already settled — Otter, Fireflies, and Granola have perfected "record it, then organize it," and there is little room left for newcomers in recording alone. Anthropic is betting that teams will pay for "the meeting ends with usable docs and whiteboards," even if only Google Meet works today. The free artifacts through February 10, 2027 are both a trial period for teams and a catch-up period for Anthropic itself — to add platform coverage, more languages, and finer-grained permissions for outside guests.
The boundary with articles already on the site also needs stating. Last cycle, cycle194 published a piece on Anthropic's self-hosted Claude Managed Agents — about agents moving into a company's own servers; this piece is about Claude walking into video calls. One is about deployment, the other about a usage scenario, so they do not overlap. Earlier Claude articles on the site mostly covered the models themselves; this one covers a workplace capability the models have grown into, again a different topic. One caveat: every fact in this article comes from Anthropic's official blog post of November 12, 2026 and third-party summaries from the same week. The author has not personally tried Claude Meetings — the beta is only for Team and Enterprise plans, and a personal account sees no entry point. Everything about "what it can do" is the official account, not a hands-on conclusion; a follow-up will be written once there is a chance to test it for real.
How it differs from Copilot and Gemini meeting AI
Microsoft Copilot and Google Gemini could already join video calls, so Claude Meetings faces an immediate question: why switch? The answer comes in three parts. The first is model capability: Claude's reputation in long-document understanding and complex reasoning is what gives Anthropic the confidence to put it in a live meeting — when someone casually asks "how much did last year's event actually cost?", it has to understand, verify, and put the number in context, not reply "please see the attachment." The second is the shape of the output: Copilot's meeting summaries mostly land in Word and Teams chats, Gemini's in Google Docs, while Claude Meetings produces a combination of documents, whiteboards, and action items — the output keeps pace with the discussion, step by step.
The third point, and the easiest to underestimate, is the neutrality of "not tied to an office suite." Copilot's meeting AI works best inside the Microsoft ecosystem, Gemini's inside Google's; Claude Meetings only supports Google Meet today, which looks like the same kind of lock-in — but its output lives in claude.ai, not in Google Docs. That means if a team later switches video conferencing platforms, its meeting notes and artifacts don't have to move. For small and mid-size teams, that feeling of "my data isn't held hostage by a suite" sometimes matters more than two extra features. Of course, that is only the current picture: whether Anthropic will later launch its own office suite and tie its output ever tighter remains to be seen.
One more practical question deserves its own mention: the competition in meeting AI is shifting from "who remembers everything" to "who remembers correctly." The fuller the transcript, the more visible the errors: a misspelled name, a copied number, a deadline attached to the wrong person — any of these can destroy a team's trust in the whole set of notes. By putting "answering live" and "writing docs live" together, Claude Meetings is actually raising the bar for itself: it has to get things right on the spot, while everyone is watching. The free beta is, in a sense, tuition for that ability to "get it right live" — feeding it real meetings so that mistakes surface early and get fixed fast.
One last note: the free beta is the best trial window
Two dates in the Claude Meetings beta are worth putting on your calendar: right now, notes, transcripts, and in-call answers are free on Team and Enterprise plans; and February 10, 2027, the date through which docs, whiteboards, and other artifacts made in calls are free. Anthropic has not announced pricing after the free period. For teams, that makes now the cheapest trial window: run it on real meetings for a month, work out how much post-meeting organizing time it actually saves, and then decide whether it is worth paying for.
The way to judge it is simple: for four consecutive weeks, count the share of meetings whose pages were actually opened afterward. If that share stays below half for a long time, the output is not entering the team's workflow, and it is not worth keeping even for free; if it stays above 70%, the subscription is worth renewing at almost any price. A tool's value is not in its feature list; it is in whether anyone opens it after the meeting.
Comments (0)
Sources
Related articles
On 20 November 2026, Thinking Machines Lab released Inkling 2 Mini: a 26.4B-parameter looped transformer that runs its middle 12 layers three times per token, fits in about 20GB, and runs offline on a laptop. It leads open-weights models of its size on DeepSWE and SWE-bench Verified — with the caveats printed in the same announcement.
On 17 November 2026, Microsoft launched MAI Code 2 and Code 2 Flash: the larger model has 77 billion active parameters and a one-million-token context, averaging 63.2% across four benchmarks — close to GPT-6.1 Sol and Claude Opus 5.5 — at $0.309 per task. Read the three caveats first: promotional pricing, the lowest score of the three, and no long-horizon work.
On 11 November 2026, AWS open-sourced UAEF: adapters turn traces from LangGraph, LangChain, Strands and Bedrock AgentCore into one format, then 36 metrics across 7 dimensions score them, flagging regressions past 5%. The judge model is hardened too — scores must cite verbatim evidence, and never decide safety.