Where your credits go

1 credit = $0.001 (a tenth of a cent) of insured AI-processing cost. No hidden multipliers, no surprise fees — here's what that pays for and how subscriptions differ from ad-hoc top-ups.

What 1 credit buys

• Document workflow (PDF → structured data) — ≈ 50 credits · ≈ $0.05
• Image OCR + analysis — ≈ 80 credits · ≈ $0.08
• Short video clip processing — ≈ 400 credits · ≈ $0.40

Where the money goes

• > 99% — Claude Sonnet 4.6 on AWS Bedrock. Anthropic model tokens — input $3.60 / 1M tokens, output $18.00 / 1M tokens, insured at the published Bedrock rates.
• < 1% — AWS infrastructure. Lambda execution + S3 + DynamoDB, plus a $0.001 run-floor per workflow for ledger write + observability.
• Margin — operating margin. A single global margin knob keeps the platform sustainable. No per-feature multipliers, no usage-tier surcharges.

Cost tables and the workflow commit ledger are open in the codebase — every charge is auditable per run.

Subscribe for access tier — top up for raw compute

A subscription pays for the access tier (priority queue, larger files, included support) and bundles a monthly credit allowance. Top-up credits buy compute time at our actual cost — they don't grant tier benefits.

| Plan | Credits / month | Subscribed rate | Top-up rate | |---|---|---|---| | Free | 100 (monthly cap) | Included | — | | Pro · $16/mo | 1000 credits | $0.016 | $0.001 | | Business · $35/mo | 5000 credits | $0.007 | $0.001 |

Top-up credits are sold at cost; subscription credits cost more per credit because the monthly fee also funds the access tier. If you only need compute, top up — if you need priority queue and larger files, subscribe.