Google Labs CC Family AI Agent: The First AI That Gets Its Own Seat at the Kitchen Table

Picture a household on a Tuesday morning. One parent has the school newsletter in their inbox. The other has the dentist reminder. A teenager has the soccer schedule as a photo on their phone. The grandmother who helps with pickups has none of it. Somebody will forget something today, and the argument about whose fault it was will happen at dinner.

On September 17, 2026, Google Labs proposed a new way to handle that Tuesday. CC, the experimental agent it launched last December as a personal daily briefing tool, has been rebuilt as an agent for families and households. Up to six people can share one CC. It has its own Google account. It reads only what each person chooses to give it, and from that it produces a single shared picture of the day, writes the dates into a family calendar, fills in the permission slip, and plans dinner.

That is the pitch. Below is every part of what Google announced, how each piece works, what Google has said about privacy and control, how CC stacks up against Meta’s rival agent launched nine days earlier, and the questions the announcement leaves open. This is a long read, because the product deserves one: it is the first time a major company has designed an AI agent around a group rather than a person.

Part 1: Where CC came from

CC did not appear from nowhere. Understanding its history explains its design.

December 2025. Google Labs launched CC as a single user experiment. The idea was simple: connect Gmail, Calendar and Drive, and every morning CC would email you a briefing about your day. The name is a play on “cc” in email, the agent copied into your life. Google says it became a daily habit for many early testers.

May 2026. The personal version was folded into the Gemini app under the name Daily Brief. That gave the feature a mainstream home and freed Labs to push CC somewhere new.

September 17, 2026. Tom Shane, Senior Product Manager at Google Labs, announced the household version. He wrote that one of the top user requests was for CC to do more to help run a busy home, and that the problem is coordination across multiple accounts: school schedules, sports practices, bills and dinner, spread across several inboxes and several people.

So CC has moved in three steps: a briefing for you, a briefing inside Gemini, and now a shared agent for the people you live with. Google’s blog title for the announcement, “CC is an AI agent for families and groups,” hints at the next step: groups beyond families.

Part 2: The core idea, an AI with its own identity

Every other consumer AI assistant lives inside your account. Gemini is yours. Meta AI is yours. CC is different. Google has given it a verified Google account of its own.

This sounds like a technicality. It is actually the foundation of the whole product, for three reasons.

It gives CC a face. When CC sends the morning brief, replies in a Google Chat thread, or adds an event to the family calendar, it shows up as CC, a distinct participant, not as an anonymous system note attached to one parent’s account. Family members can talk to it the way they would talk to each other.

It solves the “whose account is it” problem. If a household AI lived in one parent’s Google account, every other member would be sharing their data with that parent, not with the tool. A separate account means each member has a direct relationship with CC, and what they share goes to CC, not to a spouse’s inbox.

It creates a clean permissions model. An account can be granted access to a Drive folder, added to a calendar, or emailed directly. Google is using the ordinary sharing mechanics its users already understand, rather than inventing a new consent system. If you know how to share a Google Doc, you know how to share something with CC.

Google adds one hard rule: CC only responds to group members, and will not take action or share information outside the group without permission.

Part 3: What CC does every day

Google organises CC’s abilities into three jobs. Here is each one in detail.

Job 1: The morning brief, now shared

Every morning CC delivers a “Your Day Ahead” brief to the household. Where the December version told one person about their day, the family version tells everyone three things:

• Who needs to be where
• What needs to be done
• What CC already completed yesterday

That third line is new and easy to miss. CC is not just reporting the day, it is reporting on its own work. If it added three events and pre-filled a form overnight, the household sees that in the brief. This is Google’s answer to the trust problem: an agent that shows its work each morning is easier to check than one that acts silently.

The brief lands in each member’s inbox, which is the point. Nobody has to open a new app to find out that the swim meet moved to Saturday.

Job 2: Calendar and task management, on autopilot

CC watches the emails, files and messages it has been given and pulls dates and to-dos out of them into a family Calendar or Task list. It then keeps those entries current as plans change. If the school sends a revised pickup time, CC updates the event rather than creating a duplicate.

This is where the shared model earns its keep. A dentist confirmation in one parent’s inbox becomes a calendar event both parents can see, without anyone forwarding anything.

Job 3: Doing the chores nobody wants

The third job is where CC crosses from organiser into worker. Google lists four examples:

• Filling out permission slips and activity registration PDFs. CC pre-fills the form using what it already knows about the family, then asks for any missing detail.
• Building school supply shopping lists from the school’s list and what the household has said it already owns.
• Drafting weekly meal plans, presumably drawing on the family restaurant and grocery preferences stored in its memory.
• Checking live drive times via the Google Maps API between back to back activities, so the brief can warn that soccer ends at 5 and piano starts at 5:15 across town.

CC can also create shared Docs and Sheets when a task calls for one, which turns a chat request into a document the whole family can edit.

Google’s demo video caption for this section reads: CC can take action, like filling out forms, with your permission. That last clause recurs throughout the announcement.

Part 4: How sharing actually works

The most detailed part of Google’s announcement is about control, and it deserves a close look because it is where household AI either wins or loses people.

Each member decides. Sharing is per person, not per household. A parent can share the school emails while a teenager shares only their sports schedule. Anyone can change their choices at any time.

Auto “cc” CC. You can designate senders whose emails always go to CC from now on: the school, the swim centre, the vet, an airline. Once a sender is on the list, you never have to forward anything from them again. Google calls this “Auto cc CC,” which is either clever or confusing depending on your tolerance for the name.

The weekly sender review. Every week, each member gets a private list of new email senders that have appeared in their inbox, with the option to add any of them to CC’s auto share list or ignore them. This is a quiet but important design choice. It means CC’s access grows only through explicit weekly decisions, not by default, and the review is private, so your spouse does not see which senders you declined to share.

Send it to CC. For one off items, you forward the email or message to CC. Google specifically mentions photos: a picture of a birthday party invitation or a printed soccer practice schedule, sent by email or Google Chat, is enough for CC to read the details and update the calendar.

Share files and calendars directly. Because CC has an account, you can share a Drive folder or individual files with it, or add it to a calendar, exactly as you would with a person.

Put together, Google has built a system where nothing reaches CC unless a named human sent it, listed the sender, or shared the file. There is no “connect your whole inbox” button.

Part 5: The two memories

CC remembers, and Google has split its memory in two.

Household memory holds things that belong to everyone: the go to grocery list, favourite family restaurants, the fact that the family car seats five.

Personal memory holds things that belong to one person: dietary preferences, a local time zone for a member who travels, individual details that the rest of the household does not need.

When CC learns something new while doing a task, it decides which memory it belongs in. Google gives the example of CC asking for missing details when filling a form, then updating its group memory so it can be more helpful next time.

The split matters for privacy and for accuracy. A meal plan should respect one member’s allergy without broadcasting it in the shared brief. A shared grocery list should not be polluted with one person’s protein powder. Two memories let CC do both.

Part 6: What is under the hood

Google was unusually specific about CC’s architecture.

One isolated cloud computer per CC. Every household’s agent runs on its own separate machine in Google’s cloud. This is the same architecture pattern Meta used for its Muse agent: isolation is how you promise that one family’s agent can never reach another family’s data.

Antigravity. CC is built on Antigravity, Google’s agentic harness originally created for developers. A harness is the scaffolding that lets a model take actions, use tools, open forms and manage multi step tasks reliably. Google is reusing its developer grade agent infrastructure for a consumer product, which suggests CC is more capable than a chat wrapper.

The latest Gemini models. Google does not name the version, only “the latest Gemini models.” Given the September 2026 timing, that points to Gemini’s current 3.x generation.

Native integrations. CC connects to Gmail, Google Chat, Docs, Sheets, Drive, Calendar, Tasks and the Google Maps API. Because it lives inside Google’s ecosystem, it does not need third party connectors to reach the tools most families already use.

Part 7: Availability and how to get it

• Region: United States only
• Age: 18 and over
• Account: personal Google account (not Workspace)
• Platforms: web and mobile
• Household size: up to six members
• Existing CC users: will receive an email in the coming days with an upgrade to the household version
• New users: join the waitlist at labs.google/cc
• Cost: Google has not mentioned pricing. Labs experiments are typically free during testing

For readers outside the US, including India, there is no announced date. CC is a Labs experiment, and Labs products often take months to expand or are folded into other Google products, as the personal version was folded into Gemini.

Part 8: CC versus Meta Muse

Nine days before Google’s announcement, Meta launched Muse, its personal AI agent. The two products are the clearest signal yet of where consumer AI is heading, and they are not the same thing.

	Google CC	Meta Muse
Built for	Households, up to six people	One person
Core job	Coordination: briefs, calendars, forms, plans	Action: shopping, booking, negotiating, buying
Spends money	Not announced	Yes, via one time Link by Stripe cards
Identity	Its own verified Google account	Custom name and avatar chosen by user
Runs on	Isolated cloud computer, Antigravity, Gemini	Isolated cloud VM, Muse Spark 1.3, Sentinel guard agent
Data sources	Gmail, Chat, Drive, Calendar, Maps, Docs, Sheets	Gmail, Calendar, Drive, Instagram, WhatsApp, Ticketmaster, OpenTable
Memory	Split into household and personal	Personal, with a “forget” command
Availability	US, 18+, waitlist	US, 18+, free tier plus paid plans

The contrast is philosophical as much as technical. Meta built an agent that acts for you in the world: it buys the furniture and sells the car. Google built an agent that organises the people you live with. Muse is a personal assistant. CC is closer to a family secretary who never sleeps.

Google’s version is also the more cautious one. CC does not have a wallet. Its riskiest action is filling in a PDF. That reflects both Google’s regulatory exposure and the fact that a household agent has to be trusted by six people, not one. The bar is higher when the youngest user’s school emails are involved.

Part 9: The questions Google has not answered

A close read of the announcement leaves several things unclear.

Who owns CC? Someone has to create the household and invite the others. Google has not said what that person can see or control that others cannot, or what happens to the household’s memory if the creator leaves, or if the household splits.

Can children use it? CC is 18 plus. But a family agent that cannot include the children it schedules around is incomplete. It is unclear whether a parent can share a child’s calendar with CC, or whether kids’ information only enters through adults’ inboxes.

What about Google’s own use of the data? Google has said CC will not share information outside the group. It has not said, in this announcement, whether household data is used to train models or personalise ads. The privacy policy for the Labs experiment will govern that, and families should read it.

What does “with your permission” mean in practice? Google repeats the phrase for every action CC takes. It has not detailed whether that is a one time consent per task type, a confirmation for every action, or something a household can dial up or down.

How far does “groups” go? The blog title says families and groups. Roommates, sports teams, small businesses and carers are obvious next steps. Google has not committed to any.

None of these are criticisms of the design, which is the most careful consumer agent launch of the year. They are the things a family should ask before inviting six people to share an AI.

Part 10: Why this launch matters beyond one product

Three things make CC significant even if it stays a Labs experiment.

It treats the household, not the user, as the unit. Every consumer AI to date has been designed for one person. The messy reality is that logistics are shared, and CC is the first mainstream attempt to model that directly. If it works, it will look obvious in hindsight.

It answers the trust question with mechanics, not promises. A separate account, per person sharing, a private weekly review, a brief that reports what the agent did, and a split memory are all concrete design decisions that limit what CC can see and do. This is a template other companies will copy.

It reuses developer infrastructure for consumers. Building CC on Antigravity, the same harness Google offers to programmers, suggests that the gap between “AI agent for engineers” and “AI agent for your mum” is closing fast. The tools that let an agent write code are the same ones that let it fill a permission slip.

For families in the US, CC is worth joining the waitlist for, if only to see the morning brief in action. For everyone else, it is a preview of what the next generation of household technology will try to be: not a speaker you shout at, but a quiet participant in the group chat who already read the school email.

Frequently Asked Questions