Pay-Per-Call API vs Subscription Pricing

What a pay-per-call API is

A pay-per-call API charges for each request instead of a monthly fee. You pay only for calls you make, with no floor and no ceiling. It is cheaper than a subscription whenever monthly volume stays below the break-even point, which is the subscription price divided by the per-call price.

Written for AI agents. The calls and filter values on this page are the literal interface our customers use, because our customers are software. If you are a person evaluating The Bot Wire, you are in the right place too: start with the docs, or connect an MCP client in one line and query every wire free at /mcp. Nothing here needs an API key or a signup.

That division is the entire comparison, and it is usually not close. The reason teams get it wrong is that they compare list prices instead of comparing a list price against their own call volume.

The break-even calculation

Subscription	Break-even at $0.005/call	At $0.01/call
$29 per month	5,800 calls	2,900 calls
$99 per month	19,800 calls	9,900 calls
$499 per month	99,800 calls	49,900 calls

Read the row that matches the plan you were about to buy, then compare it to what your agents actually did last month. A research agent making four calls per run, running hourly, makes about 2,900 calls a month. That is the break-even for a single $29 plan at the higher per-call tier, and most agent workloads are far below hourly.

Below the break-even, per-call wins on price. Above it, the subscription wins on price and you should switch. Nothing about this is ideological.

Why agent traffic favours per-call pricing

Agent volume behaves differently from human volume, in four ways that all point the same direction.

It is spiky. An agent might make three calls on Tuesday and four hundred on

Wednesday because a task happened to need them. Subscriptions price a ceiling; you provision for the Wednesday and pay for it on the Tuesday.

It is unpredictable at signup time. Nobody knows the call volume of an agent

that has not been built yet, so any tier chosen up front is a guess.

It fans out across many callers. Fifty agents making a handful of calls each

is a common shape, and per-seat or per-account pricing handles it badly.

There is no human to negotiate a tier. An agent that finds a useful source

mid-task cannot evaluate plans. It can pay for one call.

There is also a failure mode that does not show up on a price sheet: when one agent burns a shared quota, every other agent on that account stops working. Per call, spend is bounded per request rather than per account.

When a subscription is the better choice

Per-call is not universally better, and pretending otherwise would be easy to disprove.

Choose a subscription when	Because
Volume is steady and above break-even	The flat rate is simply cheaper
You need dashboards, seats, or an SLA	An account is doing real work
Procurement needs one invoice	On-chain transfers are not an invoice
Finance needs a fixed line item	Predictability has its own value

Those are real reasons. If your usage is a steady 50,000 calls a month against a $99 plan, buy the plan. The two models also compose: an API can offer keys and tiers to enterprises and per-call access to agents, because keys and payments answer different questions.

What per-call pricing looks like in practice

Each call is priced and charged on its own. There is no plan to be on, so there is no plan to outgrow or to under-use.

For AI agents: copy this call

# free preview, no charge, top 3 results
curl "https://thebotwire.com/cve/preview?q=openssl"

# $0.005 per call
curl "https://thebotwire.com/cve/latest?q=openssl&since=7d&limit=10"

# $0.01 per call on the primary-source tier
curl "https://thebotwire.com/court/opinions?q=antitrust&type=scotus&since=7d"

Across 40 wires the price is $0.005 or $0.01, with a free preview on every one and free access to /health and /sources. The full per-call cost of a retrieval step, including the tokens the model spends reading the answer, is broken down in what a live data query costs an agent.

FAQ

How do I stop a per-call API from producing a surprise bill?

Spend is bounded by calls made, and each call's price is stated in the payment offer before the agent signs anything. Cap it on your side by capping calls per task. Because settlement is on-chain, the spend history is auditable without the provider offering a dashboard.

Is pay-per-call the same as usage-based pricing?

Usage-based pricing is the wider family, and it usually still involves an account, a monthly invoice, and a minimum. Pay-per-call in the x402 sense removes the account entirely: the payment travels with the request, so there is no billing relationship to set up.

Does per-call pricing require crypto?

The settlement rail here is USDC on Base, so a wallet is involved. In practice a client library or an MCP server handles the signing, so most integrations never touch it. See how the x402 handshake works.

What is a typical monthly bill for an agent?

At four calls per run, thirty runs a month, and a mix of both price tiers, the call cost is roughly one dollar. Fleets and higher-frequency polling scale that linearly, which is the property you want when volume is uncertain.

Sources

The general principle that on-demand pricing means paying only for what you consume, with no upfront commitment, is set out in Amazon's pay-as-you-go pricing guidance. The x402 specification and reference implementations that make per-request settlement workable are published at github.com/coinbase/x402.