Credits: balance, packs, consumption and refunds

Credits are a prepaid balance that pays for StackJack's AI automation features on StackJack-managed AI credentials. They translate the underlying AI model costs (tokens and active runtime) into simple dollar terms: 1 credit is worth $0.10 today. StackJack may adjust the credit value over time; your portal always shows authoritative balances and per-run costs, and the packs you see at purchase always state exactly how many credits you get for the price shown.

An organization that supplies its own Anthropic API key (BYOK) is billed by Anthropic instead, for its runs and its builder help alike — see Bring your own Anthropic key.

What consumes credits — and what doesn't

Credits fund the Automations features only.

What the guided builder charges for

The builder is a chat, and a single message from you can cost more than one model call:

• One message can become several charges. When the assistant needs to look something up mid-answer, it makes another model call, and each call is reserved and settled on its own.
• It can run real read-only tools against your connectors. To check that a tool returns what your automation needs, the builder may call it for real, against your live data. Those calls count against that connector's monthly allowance exactly like any other tool call, and they are subject to the same limits — a limit refusal comes back as an ordinary message rather than an error. Only read-only tools can be tested; a write or destructive tool is refused. Reading is not risk-free: a test call returns your real data into the builder conversation. There is a cap on how many tests one interview may run.
• With your own Anthropic key on file, the builder runs on your key. Builder turns and AI Assist suggestions are billed to your Anthropic account, exactly like your automation runs, and nothing is taken from your credit balance. If your key cannot be read when you send a message, the turn is refused with a message asking you to re-enter it — it is never quietly moved onto StackJack's own key. The live tool tests above still hit your own connectors and your own allowance either way.

Fast decisions are a separate lane, deliberately

A decision step and a smart filter ask TypeSafe AI for a short typed judgement. That is a different vendor from the one that runs your automations, so it is billed separately from everything in the table above and never folds into a run's own credit total.

• They are not billed today. Every call is recorded, and calls made before billing starts are never billed to you.
• If charging is switched on later, a decision call costs a small flat number of credits per call — not per token — and it appears in credit history on its own line naming how many calls it covers. A run's own credit total, its per-run ceiling and its refund rules are unchanged by it.
• A decision call made on your own TypeSafe key is never billed by StackJack, and your key is never used for anything else. See Fast Decisions and Your Own TypeSafe Key.
• Your own Anthropic key does not cover fast decisions. Different vendor, different key. Having one on file changes nothing here.
• A refused decision costs you nothing. If the feature is off, your organization has not turned it on, the request was malformed or no key was usable, the step is recorded as skipped and no charge can arise from it.

Where credits live

Credits are managed on the Credits tab of the Automations area (the old /credits address redirects there). Automations is on for every account by default. If Automations is missing from your sidebar, it has been switched off for your organization — contact support.

The tab shows, top to bottom:

1. The Concurrency card — how many automation runs your organization may have going at once, how many are running, and how many are waiting for a slot. See Concurrency, Slots, and the Run Queue and Agent Runner plans and concurrency slots.
2. Three summary tiles: credit balance, Anthropic credentials mode, and last credit activity.
3. A banner when the balance cannot be loaded, or when the account is in debt.
4. Buy credits — the packs you can buy, for owners and co-owners.
5. The Anthropic credentials card, which states which credentials your runs execute against and carries the key-management controls for eligible accounts. See Bring your own Anthropic key.
6. Recent transactions.

If your organization holds an Agent Runner plan with no Anthropic key on file, a prompt to add one sits above the tabs on the whole Automations page.

The Credits tab groups the current balance, one-time credit packs, Anthropic key mode, and recent transactions. The Concurrency card that now leads the tab is newer than this screenshot.

If a section fails to load, it shows its own error banner with a Retry button and a support code: SJ-BAL-* (balance), SJ-PK-* (pack catalog), SJ-TX-* (transactions). These are almost always transient — retry, and contact support with the code if one persists.

How a run is charged: reserve, then reconcile

Every metered run follows a reserve-then-reconcile cycle. The reservation must fit the current balance before work begins, while final reconciliation still settles the best final usage StackJack can obtain if the estimate was low:

1. Reserve. Before the run starts, credits are reserved atomically: the lesser of a runtime-based estimate (derived from the automation's model and runtime cap) and the automation's Max credits per run setting.
2. Refuse cleanly if short. The whole reservation has to fit. A balance that is positive but smaller than the reservation is not enough: the run is refused before it starts, recorded as CreditExhausted with the required and available amounts in its error message, and nothing is charged. So a run can be refused while your balance is above zero — refused up front rather than stopping halfway on a surprise.
3. Meter during the run. Usage is tracked live. If it reaches the per-run cap, the run is stopped at once and ends as CreditExhausted; the agent is not asked to wrap up first, and its session is shut down.
4. Reconcile at the end. The actual cost is compared to the reservation:
   • Used less than reserved → the difference is refunded (you'll see a Refund transaction).
   • Used more than reserved → the difference is deducted as an additional finalize transaction.

Wizard turns and AI Assist suggestions reserve and settle the same way, so the same whole-reservation rule applies to them.

Every reservation and reconciliation appears as a separate line in your credit transaction history. The ledger retains the run id for agent-run entries, although the current Recent transactions table shows the transaction description rather than a clickable run link.

What drives the cost of a run

A run's credit cost is computed from two components:

• Tokens — input, output, cache-read, and cache-write tokens, at rates that depend on the model the automation uses. Larger models cost more per token; cache reads are heavily discounted relative to fresh input.
• Active compute time — a small charge proportional to the model's actual working time (active seconds), not wall-clock time.

To see why a particular run cost what it did, open the run and read its Cost breakdown card — it splits the charge across input, output, cache-read and cache-write tokens plus session runtime. Cache traffic is usually the largest line by a wide margin. See Run Detail.

When a run ends, StackJack reads its final Anthropic usage and settles that against the reservation. If that final read is unavailable, for example after an interruption, the run is charged only for the usage StackJack can confirm, which can be less than it actually used. While a run is active, the portal's usage and credit figures are partial; a run that just ended can briefly await its final settlement. The run detail page identifies the final StackJack ledger charge once settlement completes, and separately re-prices token anatomy using the automation's current model and today's rates. For a customer-visible run total, use Billed on the run detail page.

The per-run cap stops a run

For StackJack-managed execution, the Max credits per run guardrail (settable per automation, 0.01–500 credits, default 50) is strict:

• Mid-run, reaching the cap stops the run at once as CreditExhausted, with no wrap-up turn. The check counts the whole run: its AI thinking steps before the agent starts and any earlier attempt of the same run.
• At reconciliation, the run is billed for the credits it actually used, including the model turn that crossed the cap, so a run usually settles slightly above the cap.
• If the platform restarts while a run is in progress, the run can go further past the cap before the check resumes, and it is charged for the credits it used.

BYOK runs skip StackJack credit metering, the reservation and this guardrail entirely; use Max runtime to bound them. The runtime guardrail works the same way for time: Max runtime is settable per automation between 30 and 3,600 seconds (default 300). See Guardrails and safety controls.

Refunds, by what happened

A run recorded as CreditExhausted does not restart itself after you top up. That status is terminal: the run is finished, with the reason on the run and a failure notification sent. Buying credits lets the next run start — launch it again yourself, or wait for the automation's next scheduled occurrence or trigger; either one runs normally once the balance covers a full reservation. That is different from a run that is only Queued: queued work is still waiting for a concurrency slot and starts on its own when one frees, without you doing anything. See Concurrency, Slots, and the Run Queue.

Recovered runs. On a recovered run, the run detail page reports the automatic-restart count and says that the original hold was reused. Read that note together with the run's reservation, refund and finalize entries: for StackJack-managed credits, recovery creates no second reservation and finalization is idempotent per run id. That single-StackJack-bill guarantee does not apply to BYOK: BYOK bypasses the StackJack credit ledger, and the stranded attempt and the replacement attempt may both have generated usage billed directly by Anthropic. Recovery is also a fresh execution, not a continuation: connector-side effects completed before the interruption can be repeated. Operations that must happen only once need vendor-side idempotency or deduplication, and you should inspect vendor state before manually running the payload again.

Buying credits

Only the primary owner and co-owners can buy credits (other members see a note asking them to contact their owner). Purchases are one-time — there is no credits subscription — and credits never expire.

1. Open Automations → Credits.
2. In Buy credits, pick a pack — each card shows the pack name, the number of credits, and the price.
3. Click Buy. A secure checkout opens in a new tab (Paddle, StackJack's payment provider and the same provider as plan upgrades; your billing profile is created automatically on first purchase).
4. Complete payment. Credits land on your balance automatically once the payment is processed — refresh the tab to see the updated balance and a Purchase entry in your transaction history.

Things to know:

• The pack terms you clicked are locked in at that moment: even if the catalog changes between your click and the payment completing, you receive exactly the credits and price you saw.
• If a pack shows "Checkout not yet configured", it isn't purchasable yet — pick another pack or contact support.
• If the portal cannot safely link the checkout to your organization before opening it, it refuses to open the checkout and asks you to retry — this protects you from a payment whose credits could fail to land. The message includes a transaction reference for support.

Transaction history

The Recent transactions table shows your last 20 credit movements: date, type, description, the signed amount (green for additions, red for deductions, to four decimal places), and the balance after each. You'll see purchases, run reservations and their finalizations, refunds of unused reservations, wizard-turn charges, and any adjustments applied by StackJack support. Agent-run entries are keyed to their run in the ledger; wizard turns and field suggestions use their own turn identifiers.

For a BYOK run, StackJack writes a zero-amount audit entry only after final settlement completes. Its description contains StackJack's approximate usage calculation, not an Anthropic invoice; an entry can therefore arrive after the run first becomes terminal, and Anthropic's bill remains authoritative.

If your balance goes negative (debt)

Reconciliation records the true cost even when a run under-reserved, which can push the balance slightly negative in rare cases. When that happens:

• The balance is shown in red with an Account in debt alert, and a banner explains the debt and links straight to the credit packs.
• New StackJack-managed runs are refused until the available balance can cover the full reservation each new run requests — not merely until the balance becomes positive. BYOK execution has no StackJack reservation and is not blocked by the StackJack credit balance.
• You can clear this yourself — no support ticket needed. The Buy buttons stay available, and buying a pack is how you get out: the purchase lands as an ordinary credit addition, so it settles the debt first. Once the remaining balance can cover a new run's reservation, trigger that run again. A run already recorded as CreditExhausted is terminal and never resumes after the purchase.
• If you want to understand how the overage happened, contact support@stackjack.io — but you do not have to wait for support before buying credits and starting a new eligible run.

Bypassing credits with your own Anthropic key

Eligible accounts can bypass credits entirely by supplying their own Anthropic API key — see Bring your own Anthropic key. One caveat worth knowing before you rely on it: if your key was enrolled because you hold an Agent Runner plan and that plan ends, StackJack pauses your automation runs rather than moving them back onto credits. Removing the key is what returns you to credits.

Related pages

• Bring your own Anthropic key (BYOK)
• Agent Runner plans and concurrency slots
• Run lifecycle and statuses
• Guardrails and safety controls