Skip to content

Pricing & Cost Model

How we're thinking about pricing the API once it exists — and how that relates to what it would cost to run. A fast-moving section with its own changelog. Exact numbers below are placeholders under discussion; no payment flow is live yet.

Pricing model (draft)

The direction from the strategy session: linear pay-per-use via x402, in USDC on Base, charged from request #1 — no free tier.

Why no free trial: to an autonomous agent, signing a sub-cent payment costs the same effort as signing a free auth challenge — the friction is in the process, not the price. So a free tier might buy nothing but complexity and an attack surface. Charging linearly from the first request would keep the server stateless and hyper-lean (no usage DB, no trial logic) and filter out spam automatically. Still a design choice to validate when we wire up x402.

Price per simulated gas

We're leaning toward pricing by the EVM's native compute unit — gas — not per HTTP request. The caller would supply a gasLimit; the server would quote gasLimit × base_rate in the 402 response before touching the engine, and an out-of-gas halt would protect the worker.

  • Base rate: TBD (e.g. ~$0.0000001 / gas — under discussion).
  • Deep/multi-step endpoints could use an exponential curve so expensive queries self-price out of an agent's budget.

Discounts (token)

Token holders would get a discount on the gas price, not free access — see Token Utility. Indicative tiers (under discussion): $5 → 15%, $20 → 30%, $100 → 50%. Non-holders would simply pay the base rate; zero token friction.

Future: enterprise tier

A flat annual stablecoin subscription (e.g. ~$799 USDC/yr) for budget predictability and for firms that can't hold tokens — a hybrid SaaS model to introduce "when they come."

Anti-abuse (planned defenses)

Because agents are tireless cost-optimizers, defenses would be economic and architectural rather than human-friction based:

  • Pay-from-#1 makes spam economically self-limiting.
  • Hard ceilings at the gateway (413/400) reject queries beyond a max simulation depth — no buying your way into a DoS.
  • MCP schema guardrails would instruct the agent not to issue oversized requests in the first place.
  • For any optional free/identity path: a minimum on-chain balance (e.g. ~$5 USDC/ETH) or ERC-8004 identity check defeats Sybil/empty-wallet farming. (Net conclusion from the strategy session: with linear x402, an explicit free tier likely isn't needed at all.)

Cost model (early estimate)

Revenue would need to clearly exceed the cost of compute. Main drivers we're modeling:

  • ComputeAWS EC2 for the headless API (the dominant variable cost; scales with simulation volume × gas).
  • Hosting — the website remains on a low-cost Strato V-Server.
  • Settlement — x402 facilitator fees.

The $100 discount tier is deliberately pitched near the real per-call cost level, so heavy users would be nudged toward holding the token while margins stay healthy. Detailed unit economics: to be modeled in a future round.

Changelog

Pricing Changelog
  1. v0.32026-06-30Reframed as draft pricing model — no live payment flow; conditional language throughout.
  2. v0.22026-06-30Decided: linear x402 per-gas pricing, no free tier; token = tiered discount; enterprise annual tier later. From the strategy session.
  3. v0.12026-06-30Initial scaffold — pricing/cost-model placeholders.

Add a one-line entry here whenever the pricing or cost model changes.

A living conceptualization workspace — each section carries its own micro-changelog. Latest thinking always applies.