Usage & Call Counting
No surprises — here’s exactly how we count what you do. This page walks you through how every call is recorded, what each status on your dashboard means, when a call is treated as a failure, and which calls actually show up on your bill.
How we count your calls
Every time you invoke a MicroSkill — whether through the REST API or over MCP — we record exactly one usage event on our side. The count comes straight from what actually happened to your request, so there is nothing you need to track yourself and nothing the client can nudge one way or the other.
Each event keeps the details you’d want when you look back later: which tenant and API key made the call, the skill name and version, how long it took, the meter and the number of execution credits it consumed, and a single status describing the outcome.
The big Callsnumber on your Usage page is simply the total of those events for the time window and filters you’ve picked — successful or not, every call is in there. It is a raw count of calls, not the credit total: your quota, allowance, and any overage are all measured in execution credits (see below).
What each status means
Every call settles into one of five statuses. Think of SUCCESSas the happy path — it’s the only healthy one. The other four are all flavors of failure, and knowing which is which makes troubleshooting much faster.
| Status | What happened | Failure | Metered event | Billed |
|---|---|---|---|---|
| SUCCESS | The skill ran and handed back a valid result. Exactly what you were hoping for. | No | Yes | Yes |
| VALIDATION_ERROR | Your request reached the skill, but the input didn't pass the checks (wrong shape, a missing field, or over a size limit). It still used our runtime, so it draws from your quota. | Yes | Yes | Yes |
| REJECTED | The call never even reached the skill — we stopped it at the door because of quota, a rate limit, or a missing permission/scope. | Yes | No | No |
| EXECUTION_ERROR | Something went wrong on our side or inside the skill while it was running. | Yes | No | No |
| TIMEOUT | The call ran longer than its allowed time budget and we had to stop it. | Yes | No | No |
When a call counts as a failure
We keep this refreshingly simple: if a call is anything other than SUCCESS, your dashboard counts it as a failure. So the math you’ll see is just:
Failed = Rejected + Validation Error + Execution Error + TimeoutSuccess Rate = Success ÷ Total Calls
To make those numbers actionable, the Failures card splits the total by status. That way you can quickly see whether the trouble is coming from your side (validation errors, rejections) or from ours (execution errors, timeouts).
VALIDATION_ERROR or REJECTED usually points to something in your integration or quota worth a look, while EXECUTION_ERROR and TIMEOUT are on us to sort out.What you’re billed for
When we mark a call as a metered event, we mean “this call used real skill compute, and the outcome was in your hands.” In short:
metered_event = SUCCESS + VALIDATION_ERROR
Anything we blocked before it reached the skill (REJECTED) and anything that broke on our side (EXECUTION_ERROR, TIMEOUT) is intentionally left out — you shouldn’t pay for those, and you don’t.
SUCCESS and VALIDATION_ERROR. A validation error means the request reached the skill but didn’t meet its contract (bad arguments, oversized input), and since that still spins up Squillify runtime on your behalf, it counts. Rejections and platform faults are never charged.Execution credits
Billing is measured in execution credits, not raw calls. Each metered call consumes the skill’s creditWeight — an integer that defaults to 1, so a standard call costs one credit. Volume skills publish a per-unit weight instead (pages, rows, members, megapixels) and the catalog description states the cap. A skill’s credit cost is published on its contract in the catalog.
- Calls — the raw number of requests (analytics). Shown on the Usage page.
- Execution credits — the billable unit. Your plan’s included allowance, hard cap, and any overage are all counted in credits (
SUM(creditWeight)over metered calls).
For a plan of standard (weight-1) skills, credits and metered calls line up one-to-one; they diverge for heavier flat skills and for volume skills.
A note on older data
We keep usage data for a limited retention window. If you ask for a stretch of time that reaches back further than we still hold, we’d rather be upfront than mislead you — so we mark the result as partial instead of pretending it’s the whole picture. If you need a fully complete view, narrow the time window and you’ll be back on solid ground.