pricekit

pricekit

Know what to charge — then ship it. pricekit is a pricing copilot for Dodo Payments credit billing: your coding agent reads your app, pricekit's deterministic engine recommends a pricing model and computes the numbers in the open (using the frameworks from Dodo's Pricing in the AI Age report), provisions the billing objects in test mode, hands your agent a one-line integration task, and then proves the wiring with deterministic checks.

Agents write code. Calculators do arithmetic. Pricing math is exactly the thing you never let a model improvise.

Community project — not affiliated with or endorsed by Dodo Payments.

Quickstart (~15 minutes)

In your own app's directory — no clone needed, npx fetches pricekit for you:

# .env:  DODO_PAYMENTS_API_KEY=<your TEST MODE key>

1. Plan — your agent builds the app profile (see skill/SKILL.md Step 0), then: npx pricekit plan --app myapp --profile '<json>' → THE PLAN prints: cost/action, credit price at your margin, tier, overage, rollover, promo. Every number shown. Every number overridable.
2. Provision — npx pricekit provision → meter + credit entitlement + product created in your Dodo test account (idempotent: re-runs create nothing new). Prints the one manual webhook step.
3. Integrate — paste the printed line into your coding agent: "Read skill/SKILL.md and integrate this billing into the app. When finished, run npx pricekit verify --json and fix anything red."
4. Prove — npx pricekit verify → 7 checks: env → objects → real checkout → real usage event → forged webhook rejected → signed webhook handled → balance readable. Green means wired. Not "probably wired".

Developing pricekit itself (not integrating it into an app)? Clone this repo and use npm run plan / npm run provision / npm run verify instead — same scripts, run locally.

What it provisions (trust defaults)

Object	Values
Meter	count aggregation, 1 event = 1 action
Credit entitlement	custom unit, precision 0, 30-day expiry, rollover 50% (1 mo), overage +25% invoiced at billing
Product	$tier/mo usage-based price, credits granted per cycle, low-balance alert at 10%

Everything is tagged metadata.pricekit — provisioning is idempotent and auditable. Test mode is hardcoded; pricekit refuses to touch live data regardless of the key you give it.

How the recommendation works

Ordered rules compiled from the report — first match wins, citations printed:

no paying users        → flat-simple     (Ch.10 Phase 1: "simplicity is a feature")
api-first              → usage-metered   (Q2: devs punish paying for what they don't use)
autonomous trigger     → usage-metered   (Ch.4: autonomy axis)
countable + user-run   → saas-hybrid     (Q1 + Ch.5: credits for countable value)
else                   → saas-hybrid     (Ch.10 default)
replaces human task    → note only: outcome-based "worth exploring — earn it" (Ch.7)

Blueprints

Blueprint	For	Status
saas-hybrid	chat/job-style apps	shipped
usage-metered	API-first, agent-driven	shipped
flat-simple	pre-PMF	shipped
growth-loop (referrals in credits)	—	community bounty
agent-budgets (spend-capped agents)	—	community bounty

FAQ

• Auth? Better Auth users: prefer @dodopayments/better-auth.
• Live mode? Not in v1 — deliberately. Flip your keys and re-provision when you're ready.
• Stripe? This is Dodo-native by design: the credit ledger, meters, and merchant-of-record tax handling are the point.

Roadmap

replay (re-run the math when model prices drop) · growth-loop · agent-budgets.

MIT contributors