Google CC Becomes an AI Agent for Whole Households

The AI Agent Moves Into the Kitchen

Most AI assistants are built around one person with one inbox. That is a reasonable default and a poor match for how a household actually runs, where the school email lands in one parent's inbox, the soccer schedule lives in another's calendar, and the grocery list exists on a whiteboard nobody photographs. Google Labs has rebuilt CC, its experimental personal agent, around the household instead of the individual — and the interesting design decision is not the task automation but the memory model.

• Group size: up to six household members can share a single CC
• Availability: web and mobile for US users aged 18 and over with a personal Google account, by invitation or waitlist
• Connected apps: Gmail, Google Calendar, Google Docs and Google Chat
• Architecture: each CC runs on its own isolated cloud computer, driven by Google's Antigravity agentic harness and current Gemini models

How Does a Shared AI Agent Handle Private Information?

This is the question a family agent lives or dies on, and Google's answer is a split memory. CC distinguishes between facts that belong to the household and facts that belong to one member. The go-to grocery order and the restaurant everyone agrees on are shared. A single person's dietary preference or local timezone stays attached to that person.

That distinction sounds obvious right up until you try to build it. An agent that pools everything becomes a privacy problem inside your own home; an agent that pools nothing is just six separate assistants that happen to share a logo. Drawing the line at the memory layer rather than the permissions layer is a more ambitious choice than a simple shared-folder model, and it is the part of this release most worth watching over the next few months.

What CC Can Do for a Family

The task list is deliberately mundane, which is the right instinct for a product like this. CC produces a shared daily brief, watches connected inboxes for dates worth tracking, and turns them into calendar entries and to-dos the whole group can see.

Beyond the summarising, it takes actions. Google's announcement describes CC pre-filling activity registration and permission-slip PDFs, checking live drive times through the Maps API between back-to-back activities so the schedule it proposes is physically possible, and spinning up shared Docs and Sheets for things like meal plans. None of that is technically dazzling on its own. The value is that it is the exact set of small logistical chores that currently gets done badly at 11pm by whichever adult remembered.

Who Can Use It Today

Access is limited. CC is a Google Labs experiment, currently open to US users with personal Google accounts who are 18 or older; existing CC users get an upgrade invitation by email, and everyone else joins a waitlist. Google says Canada, Australia and New Zealand will get waitlist access when the product reaches those countries.

One real gap is worth noting without overstating it: school-issued Google accounts are not supported, which is a little awkward given how much of the school correspondence a family agent would want to read arrives through exactly those accounts. Google has not said whether that changes, and for a Labs experiment the cautious scope is defensible.

Where This Sits in Google's Agent Push

CC is the consumer edge of the same agentic build-out we have been tracking through Dreambeans' US rollout from Google Labs and the real-time voice work in Gemini 3.8 Live. The through-line is Google pushing Gemini out of the chat box and into products that take actions on a schedule rather than waiting to be prompted. Follow more agent releases in our AI coverage.

The household is a genuinely hard target for an AI agent — messy, multi-person, full of context nobody wrote down. If the shared-memory model holds up, it is a more useful proving ground than another single-user chatbot.

Sources: Google Blog — September 17, 2026; TechCrunch — September 18, 2026.