agenttoolworks

Pricing

Pay for the calls you make. Nothing else.

No subscription, no seats, no minimum. 100 credits free without a card, then prepaid packs from $10 that never expire, shared across every server.

Packs

Starter

$10

1,000 credits. Enough to verify 1,000 companies.

Growth

$45

5,000 credits. 10 percent off the Starter rate.

Scale

$80

10,000 credits. 20 percent off the Starter rate.

Prices are in US dollars, charged once. Buying more lowers the rate per credit rather than unlocking features: every tool is available on the free tier.

What each call costs

A standard call costs 1 credit. A light read, meaning one that does no network work or returns a small fixed answer, costs 0.2. Reading your own usage is free.

VerifyDesk

ToolCreditsAt the Starter rate
verify_company1$0.010
verify_vat0.2$0.002
validate_iban0.2$0.002
screen_sanctions1$0.010
enrich_company1$0.010
get_usagefreefree

JobsRadar

ToolCreditsAt the Starter rate
search_jobs1$0.010
get_company_jobs1$0.010
list_supported_companies0.2$0.002
get_usagefreefree

InvoiceForge

ToolCreditsAt the Starter rate
generate_invoice1$0.010
validate_invoice1$0.010
extract_invoice1$0.010
describe_coverage0.2$0.002

How billing actually works

Credits are stored as integers, in thousandths of a credit, so a 0.2 credit call is 200 units rather than a float that drifts. You will never lose or gain a fraction of a cent to rounding, and the arithmetic is the same on our side and yours.

Every call is settled in a single transaction that writes the debit and its receipt together. They cannot disagree, because there is no moment at which only one of them exists. Every credit added to your account is written to an append-only ledger at the same time, so your balance can always be recomputed from its history rather than simply trusted. We check that every night.

A call that fails on our side is billed zero. That is not a refund policy, it is the design: there is no charge to reverse, and therefore no reversal that can fail to happen.

Because settlement happens after the work, a burst of calls started at the same moment can take a balance slightly below zero, bounded by the rate limit. We would rather carry that than bill you for work we failed to deliver.

Why usage pricing and not a plan

An agent's workload is spiky by nature. It verifies four hundred suppliers in an afternoon and then nothing for a fortnight. A monthly plan makes you guess that shape in advance and pay for the guess, and every plan boundary becomes a reason to not call the tool. Paying per call means the cost follows the work.

It also means we cannot hide behind a subscription. If the tool stops being useful you stop spending, immediately and without cancelling anything. That is a sharper incentive for us than any retention dashboard.

Questions people actually ask

Is there a subscription?
No. You buy credits when you want them and spend them when you want to. There is no monthly fee, no seat count and no minimum commitment.
Do credits expire?
No. A prepaid pack stays on your balance until you spend it, however long that takes.
What does one credit cost?
One cent at the Starter rate, and eight tenths of a cent on the Scale pack. A standard call costs 1 credit and a light read costs 0.2, so a VAT check costs about two tenths of a cent.
Am I charged when a call fails?
No. Billing settles after the work rather than before it, so a call that errors on our side costs zero credits. It still appears in your usage with a cost of zero so you can see it happened.
Do I need a card for the free tier?
No. An email address gets you 100 credits and a working key. You are never charged automatically, because credit packs are one-off purchases you choose to make.
Is the balance shared across servers?
Yes. VerifyDesk, JobsRadar and InvoiceForge draw on one account and one balance, so a key you add once works everywhere and you are not juggling three subscriptions.
How do I know what a run will cost before I make it?
Call get_usage, which is free and unmetered. It returns the remaining balance and the last 30 days of usage, so an agent can budget before starting an expensive loop.
What are the rate limits?
60 calls per minute per key by default. Exceeding it returns HTTP 429 with a Retry-After header rather than failing silently, so a client can back off correctly.