Spaces

Spaces is a desktop app that gives a team one shared space per project, where the people on the team and their AI agents work together. Instead of each person holding a private AI chat history, a space keeps the project's chats, files, and routines in one place, so Monday's decision is still there on Thursday for everyone. It is built for teams that already use AI assistants every day and want that work to be shared rather than scattered across separate laptops and browser tabs. Spaces runs as a desktop application on Apple silicon Macs and Windows PCs, and a single person can start with it free before inviting anyone else. The problem the product sets out to solve is described on its own site as five people with five private chat histories. Everyone on a team has their own AI thread, and none of them can see each other's. The project's context ends up spread across browser tabs on separate machines, so every new question begins by re-explaining the job to the assistant. Each individual gets faster, but the team as a whole does not. A shared space is the product's answer to that gap: one project, one conversation, with chats and files living in the space rather than in somebody's personal thread. Getting started follows three explicit steps. First you download Spaces; it is free, needs no account, and installs like any other desktop app. Next you connect your AI by pasting the ChatGPT, Claude, or Gemini key you already have — one provider is enough to begin with. Then you open a space by naming the project and putting a specialist to work, and from that moment the space starts holding the context. Because each person's agents, API keys, and connected accounts stay on their own computer, the account you connect is not moved into someone else's cloud. The shared space is the core of the product. Chats and files live in the space itself, not in one person's thread, so a Product Launch space can show a press list for launch week, waitlist copy, and a battery claims comparison beside a launch brief, all visible to everyone in the space. The example space is managed by a Launch Lead and also includes a Copywriter and a Researcher. Because the space holds the history, nobody has to reconstruct what was decided or re-feed the same background into a fresh chat. Rather than one general assistant, a space can hold specialists. The site frames this as hiring agents, not one assistant: a researcher, a copywriter, a launch lead, each keeping its own notes so anyone in the space can put them to work. Each specialist gets its own instructions, memory, and tools, so it can research, draft, work through an inbox, or run a report. Agents have skills and tools attached, and the studio view lets you pick an agent to chat with or add someone new to the space. Routines are the part of Spaces that keeps working when nobody is watching. They are described as things that run on a schedule — check-ins, follow-ups, and reminders. In the product demo there are three routines in the Product Launch space: a launch morning brief that runs daily at 9:00 AM, a Friday recap that runs at 4:00 PM, and a paused waitlist follow-up. The stated behaviour is that the space does the standing work and pings someone only when it needs a decision, which means routine output does not depend on a person remembering to ask. Spaces takes the position that teams should use the models they already pay for. You connect ChatGPT, Claude, or Gemini, or run a model on the computer itself — Ollama is shown running locally — and the keys stay on that person's computer. Providers can be switched per agent at any time, and more than one provider can be brought in. The stated benefit is that you are buying the space, not another subscription for tokens: Spaces is not a reseller of AI, it is the app those models work in. Agents can also use the real web through a built-in browser. Research, clicking through, and staying signed in all happen inside the app, so the agent sees the page the user sees. The site illustrates this with a supplier specification page for a solar backpack being checked against claims held in the space's files — a 65W peak panel, a 20,000 mAh pack, and a weatherproof shell — along with the instruction not to claim that the pack charges a laptop in an hour. Having the browser inside the app keeps the agent's view and the team's files in the same workspace. Spaces separates what stays local from what the cloud carries. Each person's agents, API keys, and connected accounts stay on their own computer and never go to the cloud. The cloud carries the space, the chats, the files, and the routines, and nothing else, and shared-space content is encrypted at rest with a per-space key held in Google Cloud KMS; only members of the space receive it. A space begins as yours, and when you invite someone it becomes a shared space you both work in — the same chats, the same files, the same routines, with their agents alongside yours. The outcomes stated on the site follow from that structure. Context stops being personal: the project's history, decisions, and files belong to the space and remain available to everyone in it. Standing work continues without anyone prompting it, because routines run on a schedule. Each person keeps control of their own keys and accounts, so joining a shared space does not mean handing a colleague access to a personal AI account. And because the team is not buying tokens from Spaces, adding a seat does not add an AI subscription — the models are the ones the team already pays for. Concrete scenarios appear throughout the site. A product launch team keeps a press list for launch week, drafts waitlist copy, and compares battery claims against a launch brief with a researcher, all inside one space. The same space runs a morning brief at 9:00 AM and a Friday recap at 4:00 PM, and it can follow up on a waitlist when someone switches that routine on. Agents use the built-in browser to check supplier specification pages against internal claims. A solo user can run unlimited spaces on their own machine with specialists, routines, and playbooks, and only moves to a shared cloud space when other people need to work in it. Pricing is stated clearly. The desktop app is free on your own computer with no account, and includes unlimited local spaces, specialists, routines, and playbooks, plus your own AI keys. Spaces Cloud is listed at $10.99 a year per person, described on the pricing panel as $0.92 a month, and the FAQ gives $1.49 a month or $10.99 a year per person; it adds cloud-hosted spaces with shared chats, files, and routines, plus the ability to invite anyone who has a seat. A team of five is $55 a year. Everyone who works in a shared space needs their own seat because each person runs their own agents, and anyone who works alone stays free. Enterprise is a conversation rather than a checkout, and puts Spaces inside your own cloud so the data never leaves it, with help setting the whole thing up. The desktop app runs on Apple silicon Macs and Windows PCs. The takeaway Spaces promotes is that AI work on a team should live in a place the team can see. One space per project holds the chats, files, and routines; specialists take on defined roles; routines keep the standing work moving; and the models doing the work are the ones you already pay for. Start free on your own computer, connect a key, and invite the team when you are ready.