Keypra bills on actual usage: 1 credit = 1,000 tokens. You are charged for the tokens the AI actually processes, never a rounded-up estimate. This page lists every action that uses credits, what it typically costs, and — just as important — what is always free.
Included dictation
Every signed-in user has five minutes of dictation per UTC day, shared across the creator tools and AI Coach. Record up to two minutes at a time. Dictation does not deduct your credits; Keypra covers transcription within this allowance. Generating a draft or sending a Coach message still uses its normal credits. When the allowance is used, you can keep typing; it resets at 00:00 UTC.
Actions that use credits
Typical costs below assume a standard model. Faster "budget" models cost less; larger models cost more. You always see an estimate before you run anything.
Prompt Analysis (Practice Lab)
A full rubric-based review of your prompt. Typically ~16 credits — the most thorough single action in Keypra.
AI Coach replies
A coaching conversation turn in Practice Lab, lessons or workflows. Typically ~2 credits per reply.
Workflow runs
Each step in a workflow is billed on the tokens it actually uses. A typical step runs ~5 credits; you see the estimate before starting.
Agent Builder
Creating a workplace assistant — the questions and the draft together — typically ~5–8 credits. Coach fine-tuning and "Try it" replies ~3 credits each.
Studio generations
Constraint, Format, Example and Persona Studio outputs. Typically ~1–4 credits depending on size; research-enhanced generation ~8.
Prompt enhancement
A focused improvement pass on an existing prompt. Typically ~2 credits.
Card suggestions
AI-suggested Context Cards for your prompt. Typically ~3 credits.
Toolkit generation
Personalised onboarding toolkit. Typically ~8 credits, once.
Data export
Exporting your data. ~1 credit.
Never charged
These always cost exactly zero credits:
Failed requests
If an AI request fails before the model does any work, no credits are used. You only ever pay for work that actually happened.
Pre-flight checks
The balance check that runs before an action starts is free — it exists to stop you spending credits you don't have.
Reading, writing and organising
Documents, Context Cards, Personas, lessons and browsing the app never touch your balance.
The Token Counter
Counting tokens happens on your device. Check lengths as often as you like — it costs nothing.
Cost estimates
The "estimated cost" preview shown before an action is itself free, always.
Your Settings → Analytics page shows your daily credit usage, and every AI action shows an estimated cost before you confirm it. If a workflow ever completes without the expected charge, we flag it and fix it — you will never be billed retroactively for our mistake.
How the maths works
Each action has a base cost derived from typical token use, multiplied by a factor for the model you choose. Budget models multiply by less than 1; frontier models by more. The final charge is based on the actual tokens consumed, so a short answer costs less than the estimate, never more than the tokens really used.