beta sevenk.ai runs in beta, and that covers this document too: prices and terms change as capacity grows. The version block below states what is in force today.

Per-Seat Pricing

Version 1.1 · Effective 2 October 2026

1. Flat Per Seat, Unlimited Tokens

Every seat pays one monthly price and may use unlimited tokens per day. There are no per-token charges, no overage bills, and no credit packs. The only limit is the physical one: what the serving hardware can do, and how busy everyone else is at the same moment.

What "unlimited" means: unlimited quantity of tokens per day, not unlimited speed. Your rate at any moment is your fair share of whatever the service can do then, ordered by the fair-use governor in §4: burst-class speed on a quiet night, modest speed at peak. Nobody is throttled for "using too much"; the service simply does what it can, for everyone, equally.

2. Why Per Seat

Per-token billing and zero retention cannot live together: invoicing by token means persisting per-user usage records joined to requests, and the policy is built on never keeping those. A kilowatt-hour also can't be attributed to the seat whose tokens drew it, so a per-token price for a shared service is fiction regardless. A seat stores exactly one fact, the subscription, and it prices what you are actually buying: instant, private, always-on access at interactive speed. And it points the operator's incentives the right way: the only way to earn more is to make the service faster, not to squeeze tenants.

3. The Price, and What a Seat Is

$25 per seat per month. A seat is one named person and one API key; the same person may use that key from several devices. Sharing a key between people forfeits the seat without refund, because shared keys turn capacity planning into guesswork.

There is one tier and everyone is in it: the same models, the same concurrency allowance, the same support. Price changes require 60 days' notice, never apply retroactively, and existing seats are grandfathered for six months after any increase. The number of active seats is managed against measured capacity so the experience that justifies the price holds (§5).

4. The Fair-Use Governor

"Unlimited" is only honest if congestion is handled by queueing rather than by billing:

Each seat keeps one generation in flight; a second request queues behind it, and a seat holding around three queued requests gets an immediate 429 with a Retry-After instead of an invisible backlog.
The queue is round-robin across seats, weighted so a seat that arrives after being idle takes the next slot rather than sitting behind someone's long batch.
There is no overage, ever. Load is shed with queues and 429s; no seat is quietly slowed to punish heavy use, and a heavy user simply finishes more work at the same speed as everyone else.
Abuse is not the same as heavy use: rotating keys or running a script farm on a personal seat multiplies one seat's share and is a Policy §5 violation. The remedy is suspension, never a surprise bill.
There is no daily token budget and no hidden ceiling. The soft limit is the service's tokens per second, visible in the anonymous counters on the landing page.

5. Capacity Discipline

Serving hardware has a hard throughput ceiling, so seats are sold against measured capacity rather than against infinite demand. An advisory seat cap, published and sized from observed concurrency, is reached before decode rates fall below interactive speed, and the live anonymous utilization counters show what the service is doing, so buying a seat is understood as taking a share of a genuinely scarce resource. Each capacity milestone on the SLA roadmap adds hardware, raises the cap, and triggers a pricing review: more hardware making more service available is the only good reason to sell more seats, and the only honest lever that should ever move the price.

6. The Priority Seat: $50 per Seat per Month

The one paid difference is queue position, not capability: a priority seat runs two generations in flight with double queue weight, so it is picked up first at peak. Same models, same privacy, same unlimited tokens. Priority seats are capped at a third of active seats so standard seats keep a fair experience when it is busy.

7. Billing Terms

Billing is monthly, prepaid, and auto-renewing; cancel any time at the end of a cycle, and annual prepay carries two months free. SLA credits (SLA §4) apply against the seat's monthly fee, because the seat is the unit of both the service and its remedy. Suspension without refund happens only for Policy §5 violations. Outside SLA credits there are no refunds, but there is no lock-in either: cancel and come back next month, because we hold no data worth migrating in either direction.