What one credit buys, who owns the credit pool, the difference between plan credits and extra credits, which jobs spend credits and where to track usage.
A credit is the unit of AI usage in {{brand.name}}. When an AI step writes a reply to your customer, it spends credits in proportion to how much text it read and wrote; the This month row on the Plan tab shows that month's total.
What one credit buys
One credit means a set number of tokens. A token is the piece an AI model splits text into; roughly a little smaller than a word. How many tokens make one credit is a platform setting and is the same for every plan.
Two parts enter the count:
Input — the text you give the model: the instruction, the person's message, the details coming from the flow.
Output — the reply the model produces.
The two are added together and rounded up when converted to credits. That is why even a very short reply costs at least 1 credit. Keeping instructions from growing bloated, feeding short facts instead of long passages from contact fields, and capping reply length all cut consumption directly.
⚠️ Note: When you use the platform's AI, the tokens spent are charged as credits even if the call returns an error. The provider has already billed that request. So failed attempts are written to the This month row too.
Who owns the credit pool
The credit pool belongs to the user, not the portfolio. Every account in every portfolio you own spends from the same pool. With three portfolios, all three portfolios' calls land on one counter.
In practice: a heavy AI flow you build in one portfolio also eats the credits of a flow in another. Splitting credits between portfolios is not something the panel does; you manage the split yourself when designing your flows.
Plan credits and extra credits
The pool holds two kinds of credit, with different rules.
Plan credits are your plan's monthly allowance. They renew every month — monthly on a yearly plan too. The renewal day follows your subscription's anchor day, not the 1st of the month; it is written on the Plan tab as "Your plan credits renew on {date}." Unused plan credits do not carry over.
Extra credits come from packs you buy. Extra credits never expire: they are not wiped at month end and can sit for months. But they have to wait their turn: each month your plan credits are spent first, and only once those are gone does the extra balance kick in. Extra credits are consumed once and continue from wherever they were left.
That order stops extra credits going to waste. If your monthly allowance is enough on its own, the pack you bought sits untouched.
If your plan's credits are unlimited, extra credits are never used at all; the Credits tab says "Your plan has unlimited AI credits, so you don't need extra credits." and sells no packs.
Which jobs spend credits
Credits are deducted for:
The AI step in a flow — producing a reply for a person. In chat mode, every turn is a separate call.
Image generation — the steps that generate images and the Try it on step. Calls with images spend more tokens than text-only calls.
The post commentary in insights — when you generate an AI comment for a post.
Some jobs spend nothing: the Test connection button in AI settings is free; a call blocked because credits ran out never reaches the provider, so it costs nothing; and the non-AI steps of a flow (message, condition, action, wait) use no credits.
When credits run out, the AI step cannot produce a result and the logs state why; the remaining steps of the flow are unaffected. What to do about it is in AI credits have run out.
If you use your own AI keys
If you added your own Claude, Gemini or ChatGPT keys to your portfolio, the arithmetic changes:
Your own provider bills you for the model cost, and that amount is not taken from your {{brand.name}} credits.
Each successful reply deducts a fixed platform fee from your plan. That fee is a platform setting and can be zero; if you produced a reply with your own key, a blue box on the Plan tab states how many credits it took.
Calls with images add an image fee on top of that.
A failed call costs nothing. With your own key, only successful replies are charged.
So working with your own keys cuts your monthly credit consumption substantially without taking it to zero.
Where to track credits
Three screens show the same pool in different detail.
Screen
What it shows
Plan and invoices › Plan
This month's total, extra credits left, the platform/own-key split, the renewal date
Plan and invoices › Credits
Extra credits left, this month's usage and the packs on sale
Portfolio settings › Usage
That portfolio's calls row by row: time, account, model, source, tokens, credits
The Source column on Portfolio settings › Usage reads System credits or Own keys; this list is the most direct way to see which flow is spending what. The period heading also states the pool's window, so only this month's rows appear in the list.
You also receive one email when 80% of your monthly credits are spent and one when they run out; neither is repeated within the same month.
Limits
Plan credits do not carry over. Allowance unused at month end is cleared; only extra credits wait.
Plan credits do not renew during an overdue period. The window freezes at period end; see Renewal and late payment.
The credits of a portfolio whose team you are on are not yours. It spends from that portfolio owner's pool, and if that is empty the fix is the owner's.
Credits are not money. Unused credits are not refunded and credits cannot be transferred between users.
How many tokens make a credit can change. If the platform changes that ratio, new calls use the new one; past records stay as they were.
Spending credits well means calling the AI only in the steps that genuinely need it; that cuts your cost and speeds up your replies at the same time.