Agent credits and usage

The three tiles cover your whole period. The buttons underneath don't change them.

This trips up almost everyone the first time. Credits used, Days to limit and Credits remaining always describe your current credit period, whatever is selected below them; Today · This month · 7d · 30d · 90d only re-cut the charts and the breakdown table. So the card above can honestly read 492.3k used while the chart one card lower totals a few hundred. One is the period; the other is today.

• The three tiles

  Cover

  Your whole current credit period, always.

  Answer

  Am I going to run out?

  Changed by

  Your plan and your spend — not the range buttons.

• The charts and the breakdown

  Cover

  Whichever range button is selected.

  Answer

  What did we spend it on, and when?

  Changed by

  The range buttons, and the dimension tabs below them.

Credits remaining includes any credits you have bought on top of your plan, and says so — it reads incl. N paid when there are some. Days to limit is a projection from your average daily rate across the period so far. It shows — when you are not on course to hit the cap before the period resets, which is good news rather than missing data.

"This month" usually means your billing period

The wording says month; the number behaves like your subscription. Usage resets when your billing period rolls over, not on the 1st — so if your plan renews on the 12th, the figure you are reading runs the 12th to the 11th, and Days to limit is measured against that reset rather than the end of the calendar month. Trial accounts are keyed to the trial window the same way, and only accounts with neither fall back to a calendar month.

It matters when you compare notes with your invoice, or when someone asks why usage "reset early". It did not; the period rolled.

Where the credits went

By service is the view worth reading: one row per service, its credits for the selected range, and its share of the total, so a glance tells you whether your spend is ordinary agent thinking or something more expensive. Every Sydekick message and every drafted post or reply runs through a language model, so that row carries the conversational and drafting work and is where to look first.

By client, By initiator and By feature re-cut the same total by who or what it was for, which is how you find one client or one automation quietly consuming the period. By action currently labels its rows with the engine's internal action names rather than plain English, so it is the least useful of the five.

Above the table, the donut and Top clients show the same split visually, and a Playbook usage panel appears once a scheduled playbook has spent credits — the fastest way to spot a routine running more often, or more widely, than you meant it to.

What can use credits

Anything the agent does with a language model — every message in a Sydekick conversation, every drafted post, every drafted review reply — plus the work it does on your behalf: generating an image, scraping a page, running a search or an SEO lookup, sending a text or an email. Practice and rehearsal runs count too, because they run the real agent on your real data.

Reading a screen costs nothing, including this page and the audit log. Nor does testing the agent's connection — that panel says so itself: "No LLM call, no credits spent."

There is no rate card in the product. Different services cost different amounts per action, and this breakdown is how you find out what yours cost. Two places print a single figure: a skill's practice run reports This practice used N credits, and the playbook practice step warns up front that it spends some. Use those for a feel for one run before you schedule it across fifty locations.

As you get close to the limit

A banner appears on this card at 75% of the period's credits, and at 90% it becomes a warning with a Buy credits button beside it. Act there rather than waiting: by the time the cap is reached, work that was mid-flight has already been interrupted.

If you do run out, the agent stops rather than overspending. The app says you are out of agent credits, Sydekick answers "You're out of credits. Top up to keep chatting", and scheduled work is paused until your credits reset next billing cycle. Buy credits in Settings → Billing is the immediate fix — purchased credits never expire and are only drawn once plan credits run out, so buying early costs you nothing. Upgrading raises the recurring limit instead.

Cross-checking against the audit log

Agent → Audit logs reads the same ledger, so set its Date to This month and its Credits in period tile is the headline from this page. The two divide the work: this one tells you which service spent your credits, and the audit log tells you which individual action did.

One caveat if you go there to reconcile a figure. This month does not mean the same thing on both halves of that page: the tile follows your billing period, the way the numbers here do, but the rows listed beneath it are trimmed to the calendar month. On a plan that renews mid-month the two cover different spans, so read the tile as the total and the rows as a sample of it rather than trying to add the rows up to the tile.

Next

• How do I read the Agent audit log and find one action?
• How do I configure a playbook and run it now?