If your AI credit is draining faster than expected, you don't have to guess why: the Portfolio settings › Usage screen shows every call, one by one. The same screen also prints the portfolio's plan limits, so "have I used up my account allowance, how many more people can I invite?" is answered here too. This guide covers both sections and the six columns of the table.
⚠️ This screen is open to those who can change settings: the portfolio owner and its admins. It shows figures for this portfolio only; totals spanning all your portfolios (the credit pool, contact and automation counts) live on the owner's My user settings › Plan and invoices screen.
What it is for
The screen says two things:
Plan limits — how many accounts and team members you are allowed in this portfolio, and how many you have used.
AI usage — which account, which model, how many tokens and how many credits were spent this period.
They share a screen because they are two halves of the same question: what does my plan give me, and how much of it have I used?
How to reach the screen
Open the account switcher at the top left.
Click the gear icon next to the portfolio name.
Pick Usage from the menu.
Plan limits
The first section, with two rows. Its hint names the source: "Comes from the portfolio owner's plan."
Accounts. How many accounts are connected in this portfolio and how many you are allowed. Written as 2 / 5, or 2 · unlimited if there is no limit. The figure matches the badge in the header of Portfolio accounts.
Team. How many people are in this portfolio and how many seats you have. Same format. The count includes the portfolio owner and counts pending invitations; for details see Team and invitations.
Both limits are per portfolio: with three portfolios you get five accounts and five seats in each (if your plan says five). Raising a limit means changing the plan, and the plan lives on the portfolio owner's user account; see Plan and invoices.
AI usage
The second section. Under its heading the period is printed: something like Period: 1 September – 1 October. That period is the portfolio owner's credit period — not a calendar month, but the window calculated from the plan's renewal date. For details see Plan and invoices.
The two source totals
Below the period line sit two rows:
System credit — calls made from the panel's own provider account. Each row shows how many calls and how many credits were spent.
Own keys — calls made with the keys you added.
Read together, these two figures tell you whether switching to your own key worked. After adding your own key, the expected result is that the "System credit" row stops growing; if it keeps growing, your key's scope may not cover that account, or the key may not be working. See AI keys.
The call list
Under the totals sits a table, newest call first. It has six columns:
When. The date and time of the call, formatted in your interface's time zone.
Account. Which account the call was made in. The account name is the same one you see in the account switcher. If the account was deleted since, the row reads Deleted account — the usage record does not go with the account.
Model. Two lines: the provider's name on top (for example "Claude (Anthropic)") and the model actually used underneath. If you left "Default model" on your key, you see whichever model the provider treated as default that day.
Source.System credit or Own keys. Shows the source used per row.
Tokens. The total tokens spent on that call (input and output together). A token is the unit of text a model processes; a long instruction or a long customer message raises the count.
Credits. What that call deducted from your credit. On calls made with your own key this is only the panel's service fee (if one is defined); what you pay the provider doesn't show here, it is in the provider's own dashboard.
The list paginates; move through it with the Previous / Next buttons underneath.
If there was no AI usage this period, the table is replaced with "No AI usage this period."
Reading the usage
Concrete conclusions you can draw from the screen:
If one account is eating the credit, read the call list by the Account column. If one account's rows fill the list, its flows may hold an unnecessary AI step — for instance a step classifying every message when classifying only the first would do.
If token counts are high, your instructions are long. Shortening the fixed instruction in an AI step lowers the cost of every call.
If many rows share the same minute, a flow may have looped. Check that flow's session log.
If the credit column is higher than expected, the model's thinking may be on. Thinking spends tokens; you decide that per key.
A concrete example: a shop spends 2,000 credits a month and, looking at the list, sees half the rows come from one account and one model. That account's flow has a step summarising every incoming message. Putting the step behind a condition so it only runs on long messages halves the usage.
What is deliberately not here
The screen's scope is intentionally narrow. These are not here:
Your credit balance. How much is left belongs to your user account: My user settings › Plan and invoices › Credits.
The total across all your portfolios. This screen shows the open portfolio only.
Contact and automation counts. Those are in the user totals too.
Message and conversation counts. The panel's own figures are in the Reports section; this screen shows AI calls only.
Media storage. Space used is printed on the media library's own screen; see Media library.
Common mistakes
Taking the period for a calendar month. The period is the plan owner's credit period; it may not start on the 1st.
Taking the credit column for the provider's invoice. On calls with your own key, the credit here is only the panel's share.
Taking "Deleted account" rows for an error. Usage records are kept even when the account goes; this is deliberate.
Taking the limits as per user. Account and team limits are per portfolio.
Trying to change the plan from this screen. There is no plan link here; the plan lives on your user account.
Looking for the credit balance here. This screen shows spend, not what is left.
Related concepts
For your credit balance and plan changes see Plan and invoices, to add your own key see AI keys, and for the choice between system credit and your own key see Using your own AI key.