Usage limits & billing

Two mechanisms will govern how much you can call the API and who pays for it: per-key rate limits and quotas throttle a key, and a prepaid wallet funds whatever gets through.

Rate limits & monthly quotas#

When creating a key in Settings → API Keys (see Authentication), you will optionally be able to set:

• A requests-per-minute rate limit, enforced as a token bucket.
• A monthly usage cap, expressed as an amount in EUR.

Leave either unset and it stays unlimited. The monthly cap is a best-effort ceiling, not a hard, race-free limit; under heavy concurrent load a key may slightly overshoot before the check catches up. It is meant to bound runaway usage, not to meter billing precisely.

Exceeding either limit returns 429, with rate_limit_exceeded or monthly_quota_exceeded. See Errors for the exact payload.

The prepaid wallet#

Usage on /v1/responses will be funded from a single EUR wallet per billing subject: a user, or an organisation. Usage draws the wallet down as it accrues; topping up is a one-off payment from the web app.

Individuals hard-stop at €0 and are never invoiced: there is no surprise bill, the key simply stops working with insufficient funds until the wallet is topped up.

Organization billing modes#

Organisations will choose between two modes:

• Prepaid: same hard-stop as individuals, usage stops when the balance reaches €0.
• Pay-as-you-go: an admin sets a monthly spending limit; the wallet can run negative down to that limit, settled monthly against the organisation’s existing payment mandate. No new payment method is required beyond what subscriptions already use.

Insufficient funds (402)#

Once a subject’s wallet (or an overage organisation’s spending limit) is exhausted, POST /v1/responses returns 402 with wallet_insufficient_funds or spending_limit_exceeded. See Errors for the exact JSON shape of both codes.

Usage & revenue reporting#

Organisation admins will be able to see usage and spend broken down per API key, so a team sharing an organisation can tell which integration is driving cost. This is not live yet.

See Authentication for how limits are set on a key, and Errors for the exact 429/402 payloads.