AnyBot Tokens (ABT)

Every reply, rewrite, and image runs on a different AI model with its own provider pricing. An AnyBot Token (ABT) is the single, normalized unit that hides all of that. Whatever model does the work, its cost is converted to ABT and metered against your plan, so you reason about one number instead of a dozen provider price lists. Your plan sets the allowance and usage draws it down as you work, and if a pool runs out before its window resets, you can buy a token top-up without changing plan.

How metering works

Each AI call returns the underlying provider's input and output token counts. AnyBot looks up that model's rate, converts the tokens into ABT, and debits the total from the matching pool in your allowance. In flow:

Provider tokens (in / out) → model rate → ABT → pool quota → usage dashboard

Media generation is metered differently, because it isn't priced in language tokens. It's billed per unit of output: images per image, video per second, avatar/presenter per minute. A generation that fails is not billed, and a retry after a failure isn't double-charged. Only the successful run draws down the pool.

The six pools

AI work is split across 6 so heavy use in one area never starves another. Each pool has its own reset window, tied to your own billing date rather than the calendar.

Resets: your billing date, not the 1st

Windows are personal, not calendar-fixed:

• The daily assistant pool resets at midnight each day.
• The monthly pools reset on your billing date, the start of your current subscription cycle. With no active cycle, the window falls back to the anniversary of your account's creation day (clamped for short months: a 31st anchor lands on the 28th/29th in February). So "monthly" means your month, which may begin on the 14th, not the 1st.

What happens at the limit

Pools are independent. Each is checked on its own before the work runs:

• Exhausted: when a pool's used amount reaches its allowance, the next action against that pool is refused with an HTTP 429 (error E4002, a clear "limit reached for that pool" message). Nothing you've already produced is touched, and other pools keep working. Running out of image generation never stops customer conversations. The pool frees up again at its next reset.
• Not on your plan: a pool with a zero allowance means the feature isn't included in your tier (an upgrade prompt), which is distinct from "exhausted."
• Unlimited: some tiers set a pool to no cap; it never blocks.

Your dashboard shows used-versus-included per pool, a per-model breakdown so you can see which model is spending, and a single headline percent available, a weighted view across all your capped pools.

Worked example

You spend a morning building a campaign:

1. You ask BotFather to outline the campaign → debits the assistant pool (daily).
2. It drafts social copy → debits content generation (monthly).
3. You generate three images in Media Studio → image generation bills per image; one image fails, you retry, and only the successful re-run is charged.
4. All morning, customers chat with your deployed agent → each reply debits customer conversations (monthly).

Late in the day your assistant pool hits its daily ceiling: the next BotFather message returns a 429 "limit reached." Your customers are unaffected. That's a separate monthly pool, and the assistant pool is available again after midnight.

Limits & honesty

FAQ

Can I buy more ABT without upgrading? Yes. Token top-ups are sold as one-off packs, bought by the workspace's payer from the billing area. A top-up is fungible across pools, is spent only after your plan's allowance for that pool is used up, is consumed oldest-first, and survives your monthly reset. One limit matters: a top-up extends a pool your plan already grants. It never opens one your plan excludes. To raise the plan allowance itself, move up a tier on the Plans page.

Do unused tokens roll over? No. Each window starts fresh at your reset; the daily pool resets nightly, monthly pools on your billing date.

Why did a big model "cost" less ABT than a small one? ABT reflects each model's rate and how many tokens the exchange used. A short reply from an expensive model can bill less than a long reply from a cheap one.

Does a failed image still cost me? No. Failed generations aren't billed, and retrying after a failure doesn't double-charge. Only successful output draws down the pool. The usage dashboard's per-model breakdown lets you trace exactly where the rest of your allowance went.

Next

• Plans: how allowances scale by tier.
• Media billing: per-unit charging for images, video, and presenter.
• Promo & referral codes: save at checkout.
• Billing & ABT overview: how plans, pools, and codes fit together.