You buy work, not tokens

Sintora Content Units — a unit of finished work, not of computation. A post is 1 SCU, an article up to 1,500 words is 10, a landing page is 8, a competitor analysis is 20. Internally the system converts that into tokens for whichever model suits the task, but you never have to think about it: you are buying an amount of work you can plan around, not a technical quantity. There is one counter for the whole platform: the same unit is spent on a post in Marketing, a call summary in Sintora Meetings, a release note in Developer and any Synapse operation.

A run is 0.05 SCU, and an AI node firing inside it is 0.2. So twenty automation runs cost the same as one social post. A team on Operations Standard running 400 automations a week, a third of them through an AI node, spends about 44 SCU — roughly 2% of its weekly capacity. The rate is deliberately this low: most of what happens inside a run is database writes rather than a model call. What is worth charging for is the model call, not the run itself. That is the capacity a run spends, and the suite plan is what provides it. How many automations you may keep switched on is a separate question with a separate answer: it is a term of the Automations module's own plan.

Because that is what it actually costs to produce. An article carries an outline, a draft and an edit behind it — ten times the generation a post takes. When we priced it lower, articles were worth making and posts were not, although the difference should be nothing to us. Now the weight of each operation matches the work in it, and it makes no difference what you create: the unit means the same thing everywhere.

The operation price list is a promise: a post costs 1 SCU today and will cost 1 SCU a year from now. Models get cheaper, and when they do we raise the weekly allowance rather than rewrite the list. Otherwise you could not plan anything — and planning is the entire reason this unit exists.

40% of the operation price — a rework regenerates only part of the material, so charging in full for it would be wrong. Translating into one more language is 60%. Batch work from 50 units carries a 50% discount, because those jobs genuinely cost us less. And if the result was poor, regenerating the same brief is free — up to 3 times. We do not charge for our own misses.

Because a token means nothing for your work, and model prices change several times a year. Keeping that mechanism on our side means we can change model or provider without your plan moving. Transparency here is not about showing the kitchen — it is about the price not changing without warning.

A monthly limit can be burned in two days and then waited out for three weeks. A daily one is too short for the way work actually arrives: a campaign is three loud days inside two quiet weeks. A week is the unit marketing already plans in, and it has slack built in. Unused capacity rolls into the next week, up to one full allowance: a quiet week pays for a launch week.

Work is not blocked. You get a choice: draw on the rollover, buy a pack, or continue on Monday when the allowance resets. Everything already produced stays where it is — we do not switch the product off over a counter.

Nothing separate happens — it is already accounted for. AI in Operations and Developer (request triage, answers from the wiki) and in the marketplace modules (call summaries) spends SCU from the same shared pool as content in Marketing. Each suite contributes its weekly allowance to that pool, and it can be spent anywhere.