Simple, scalable pricing

You do, directly to Anthropic, at Anthropic's own published prices. We add nothing to it and never touch it. Connect your key once and every agent you own runs on your account — which is why runs are unlimited here and metered almost everywhere else. A typical run costs a few cents; most businesses spend less on a month of agents than on one afternoon of the work they replace.

Because a plan that includes it has to price for the heaviest month you might have, and you pay that every month whether you use it or not. Here the subscription buys the agents — the prompts, the guardrails, the refusals, the schemas — and the running costs what it costs. Use them twice this month and you pay for two runs.

Reading the catalogue needs nothing at all — no card, no key. Running an agent needs your own provider key, on every plan including the free one, because the model request is made against your account. Connecting one takes about a minute, and from then on your provider bills you directly for what the models cost.

One execution of one agent, whether triggered from the dashboard, the API, or a workflow node. Each run also records the input and output tokens it actually consumed.

The metered action is refused with a clear message naming the metric and the limit. Nothing is silently throttled and nothing runs at degraded quality.

No. Agents are not sold one at a time — every paid agent comes with a plan. The entry plan lets you pick any three of them; the plans above it open the whole catalogue. Your plan then governs how much you can run them.

Yes. Upgrades and downgrades are handled in the customer portal of our payment provider, reached from your billing page. They are the merchant of record and apply any proration.

Negotiated volume, custom agent development, SSO, a security review, deployment in your own cloud where required, and an SLA. Enterprise is quote-only.