The model: charge per call with x402
The cleanest way to monetize an agent-facing tool is per-call payments over x402. Instead of API keys and monthly invoices, your endpoint returns HTTP 402 with a price when called without payment; the agent pays and retries; you return the result. No accounts, no billing portal — the payment is the access.
Step 1 — Wrap your handler
Use an x402 middleware/SDK to wrap the handler you want to charge for. The wrapper returns the 402 quote automatically on unpaid requests and lets paid requests through. You set the price (often a fraction of a cent), the asset, and the pay-to identity. Most SDKs offer Express/Hono/FastAPI wrappers so this is a few lines.
Step 2 — Price it for agents, not humans
Agent pricing is different from SaaS pricing. Agents make many small calls and are extremely price-sensitive per call but indifferent to signing up. Price per call, low (sub-cent to a few cents), and make the value obvious from the endpoint description so an agent can decide to pay without a human reading docs.
Step 3 — Make it discoverable
A paid endpoint nobody can find earns nothing. Publish your resource in an x402 discovery catalog so agent frameworks and indexers can list it. Each catalog entry should carry the price, network, asset, and pay-to address in canonical x402 format. This is the difference between "an endpoint exists" and "agents can find and buy it."
Step 4 — Let buyers onboard in three lines
On the buyer side, the agent should reach a first paid call almost instantly. With a good SDK that looks like: create an identity, claim starter credits, then fetchWithPay(url) which handles any 402 automatically. If onboarding takes more than a few minutes, you lose the agent.