Work is already going through consumer chat accounts you do not administer, on terms nobody read, with no record of what was asked or what answered. The decision in front of you is not whether to adopt AI. It is whether the version your team uses is one you can see, cap and account for.
Quorum puts a question to several frontier models at once, has them deliberate, and returns one reply drawn from their work. Being right more often is why people want it. What makes it defensible is that every answer arrives with its own record.
More than one company’s model saw the prompt, so you are owed a list of which ones. Quorum returns that list on every call, in the response body, without being asked — live today, not a policy page and not a roadmap item.
A five-person startup and a five-thousand-person enterprise pay the same per-token rate, because a token does not know who sent it. What varies between them is which controls they need — shared billing, member roles, per-key limits, their own provider keys — and turning those on is self-service, not a negotiation.
Volume and commitment pricing is not published yet — prepaid, per-call billing is the only rate today. What we do not do is hand out free tokens sized to inflate a usage number, which is a discount you pay for later in a renewal conversation.
Three seats is the smallest team you can buy — a minimum quantity, not a flat fee. Three seats is $90 a month; five is $150; you add or reassign seats whenever you like. Annual is $300 per seat per year, the same two-months-free shape as the individual plans.
Every seat gets Pro-equivalent model access. Full plan comparison is on the plans page.
Written as a description of what exists rather than a feature list, because the useful thing this section can do for you is be checkable.
Membership is managed at your team page once an organisation exists.
Quorum runs on a prepaid balance. There is no invoice, no monthly surprise, and no mechanism by which a runaway script produces a bill you have to argue about.
Seat pricing is published on the plans page. Enterprise API access is prepaid and billed per call — there is no published volume rate card yet. Your organisation’s own spend figures stay behind sign-in, on your dashboard.
Stated first and without softening: a deliberation sends your prompt to every provider behind the mode’s seats, plus the provider that synthesises. That is three or more companies for one question. It is inherent to how the product works and there is no configuration that avoids it.
What we can do is never make you wonder which ones. Every response carries quorum.engines — the model families that saw that specific call. Not sampled, not aggregated monthly, not available on request. Per call, by default, in the body.
Bring-your-own-key is live and self-serve for OpenAI, Google, Groq and Bedrock. Seat traffic then runs on your provider contracts, under your provider terms, on your provider bill.
Bedrock connects through a cross-account IAM role you create, assumed per request. Quorum never holds a long-lived AWS secret of yours.
Orchestration always runs on Quorum’s own keys. The passes that judge the seats’ answers and synthesise the final reply see that content regardless of whose keys the seats used.
We would rather you learn that here than in a security review.
One blanket claim about data handling is the wrong shape for this product. What actually governs a prompt is which providers that mode seats, and that varies mode to mode. A mode profile should carry its own posture, derived from its real configuration rather than asserted.
Would be derived from the mode’s real seat configuration, the same way its ingredients already are — so a mode could not claim a posture its seats do not support.
The mechanics of quorum.engines, receipts and key setup are on the API reference.
Quorum does not hold SOC 2. There is no signed DPA on offer, no SSO or SCIM, no enterprise audit logging, and no published subprocessor list. None of these is in progress; we will start when there is dedicated enterprise demand that justifies the cost and the timeline, and we will publish the report when it is issued rather than the intention now.
If any of those is a hard requirement for you today, this is the paragraph that saves you a call, and it is deliberately placed where you will read it before you invest any more time.
Government, defence and federal certification — FedRAMP, ITAR and similar — are out of reach and we are not pursuing them. Said plainly here rather than discovered in a discovery call that does not exist.
Four steps, each of which lands on a surface that already exists. No demo, no qualification call, no waiting for someone to get back to you.
Every new org gets a $10 credit to try the platform and see if it could help their organisation deliberate better. It is on the balance when the organisation is created, so the first call runs without a card. Anything unused after 30 days is deducted.
There is no sales organisation to route you to, and pretending otherwise would waste your week. For a security questionnaire, an unusual requirement or a redline request, the address is support@quorum.dog, and it reaches the people who built this rather than a queue.
We are not going to promise a response time we have not measured. What we will do is tell you plainly whether the answer is yes, no, or not yet — which is the same standard section 05 holds itself to.
An analyst working in a spreadsheet does not want a new tab; they want a better answer in the cell they are already in. Quorum is reachable from inside tools that speak MCP — a spreadsheet assistant, an IDE, an agent framework — as six tools rather than a website.
Honest scope: a hosted remote server is now live at https://www.quorum.dog/mcp, so there is nothing to install — point any remote-MCP client at it with your API key as the bearer token. It is not yet listed in any connector directory. Wiring instructions are on the API reference.