Real questions.
Plain answers.

ChatGPT is one assistant. Space Office is a team.

Each of the 24 specialists has its own role, training, tools, and quality bar — a writer that knows your brand voice, a designer that knows your visual system, an SEO expert that knows your site. They work in parallel, hand off to each other, and Hydrogen reviews everything before it reaches you.

The bar isn't “can the model produce something?” — it's “would you actually ship this?”

It already has. All 25 agents are live and working today — Hydrogen plus 24 specialists. There's no waitlist and no staged rollout to wait through.

A further 15 agents are in development. You'll see them appear in your roster as they ship, at no extra cost beyond the usual $25/mo per agent you choose to add.

Hydrogen reads the brief, asks only for missing essentials, splits the work into specialist tasks, reviews each output, and sends back a polished handoff for your approval.

Yes. Paid workspaces run on their own AWS instance rather than a shared worker pool. Compute is billed separately at AWS cost, and you can resize when workload changes.

Honestly? Because it's a great mental model. Atomic number maps to importance — Hydrogen (1) is your manager, the most-used specialists are early in the list, the niche ones come later. You can see the whole team at a glance.

Also: every team needs names, and “Nitrogen” is more memorable than “Content Writer Bot v2.”

Hydrogen is your AI project manager. Before any specialist's work reaches you, Hydrogen reviews it against the brief, your brand voice, factual accuracy, and a quality bar.

If it's not good enough, Hydrogen sends it back with specific feedback — “the intro is weak, try the angle from your earlier draft” — and the specialist fixes it. You only see the polished version.

Hydrogen checks outputs against the brief and sends issues back for revision. You review and approve the final work. No measured accuracy or issue-detection rate is claimed.

Not perfect. AI work still benefits from a human glance. But the difference between getting first-draft AI slop and pre-reviewed work is significant — and it's most of what you're paying for.

Yes. Every task has a review log: what Hydrogen checked, what passed, what got sent back, and the revision notes. You can override Hydrogen if you disagree.

Yes. Hydrogen plus 24 specialists are live and working today — engineering, design, growth, brand, revenue, and research. Nothing on the roster is a placeholder.

A further 15 are in development and shown separately so you can see what's coming. Hydrogen will tell you if a brief needs one that isn't ready yet, and what to use instead.

You answer a few questions when you sign up — what your business is, what you're trying to do, what kind of work you need. Hydrogen recommends the two specialists your $60 base covers, matched to that. You can swap them anytime, and add more for $25/mo each.

Yes — anytime, no penalty. Add a specialist for $25/mo, remove one and the charge stops on the next cycle. Mix and match.

Every workspace includes Hydrogen and two specialists in the base plan. You can add any live specialist from the roster for $25/month each, then remove or swap specialists as priorities change.

Each specialist has its own prompt, tools, training material, and quality bar — and yes, a tone. Nitrogen handles long-form copy while Cobalt plans social distribution. Boron focuses on interface design while Carbon tests software.

It's not characters; it's specialization. The personalities are a side effect of giving each one a clear, narrow job.

$60/month for Hydrogen and two specialists of your choice. Any of the other 22 are $25/month each.

Two things sit outside that: your AWS server (from about $30/month) and your AI provider's token bill. Both are passed straight through — we never mark either up.

It was a flat rate that bundled compute into the subscription, which meant light users subsidised heavy ones.

Splitting it out drops the base to $60, lets you pay for the server size you actually need, and means the AWS and provider costs reach you at cost instead of through us.

Every workspace runs on its own dedicated AWS instance rather than a shared pool. Passing Amazon's published rate straight through means you scale the machine to your workload — and you can check what you're paying against AWS's own pricing page.

Three sizes: Light (t3.medium, ~$30/mo), Standard (t3.large, ~$60/mo), and Performance (m5.xlarge, ~$140/mo).

Your team works through briefs more slowly. It never changes the quality bar — Hydrogen reviews everything the same way on every tier.

Upgrading is one click, effective immediately, and prorated. You're never stuck with the size you started on.

On paid plans, yes. Your AI usage runs on your own account — Anthropic, OpenAI, or Google — so you pay the provider directly what they charge. We don't add a markup.

This keeps the subscription cheap — we charge for the orchestration, the specialists, and Hydrogen QA, not the tokens — and it means heavy users don't subsidise light ones. Hydrogen's QA pass is part of that usage: every output is generated and reviewed.

Depends on how much you use it. Light usage (a handful of tasks per week) is typically $5-20/mo on the AI side. Heavy usage (a busy team running content + design + SEO daily) tends to land in the $40-120/mo range.

You set spending caps directly with the AI provider. Hydrogen also tells you when you're approaching limits.

Yes. Paid plans use your own Anthropic, OpenAI, or Google account for model usage. You pay the provider directly, set your own caps, and Space Office adds no token markup.

Space Office uses isolated workspaces, dedicated compute for paid accounts, scoped connected apps, and Hydrogen review before delivery. We do not claim SOC 2, HIPAA, or ISO certification yet.

Yes. After launch we'll roll out a managed option for people who'd rather not deal with provider accounts. It'll cost more (we'll pass the AI fees through plus a small handling charge) but you won't see a bill from Anthropic.

Briefs, drafts, and outputs are stored in your Space Office account. Tasks are sent to the AI provider you connected (Anthropic, OpenAI, or Google) under their data terms — none of the major providers train on API traffic by default.

We don't sell your data. We don't train on it. Full data export available anytime.

Monthly: Cancels at the end of the current month. No fee.
Annual: Prorated refund within 30 days. After that, runs to term.

One-click cancel from your dashboard. We don't make you call anyone.

Yes — annual takes two months free off the subscription, and that applies to the base and to every agent you've added. A $60 base works out to $600 for the year.

Compute is the exception: it's an AWS passthrough billed monthly at their rate, so there's nothing for us to discount there.

Small teams that want to do more without hiring more — agencies, consultancies, solo founders, freelancers scaling up, lean ops teams.

If you've ever thought “I just need someone to draft this and someone to design that and someone to make sure it ships” — that's the shape of this.

If you're a developer who wants raw model access to build something custom — go straight to the AI providers, you don't need us.

If you need regulated-industry compliance (HIPAA, FedRAMP, deep audit trails), we're not there yet.

If you want a single chatbot interface — ChatGPT is fine, you don't need a team.

No, and we don't want to oversell that. AI specialists are great at volume work — drafts, first passes, research, ops. They're not great at decisions that require judgment, taste calls only you can make, or relationships with humans.

Most of our early customers use Space Office as extra hands: the specialists do the first 70%, the human team does the final 30%. That's where it works best.

Multi-seat shared accounts ship in the first quarter after launch. For now, one workspace per subscription, with the ability to invite reviewers (read + comment, no billing access).