> ## Documentation Index
> Fetch the complete documentation index at: https://docs.quotamint.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Limits and concurrency

> Design around request, metadata, rate, and accounting guarantees

## Request limits

* `customerId`: 255 characters
* `feature` and `event`: 120 characters
* `quantity`: integer from 1 to 1,000,000
* `idempotencyKey`: 255 characters
* Request body: 64 KiB by default
* Metadata: JSON object with plan-specific byte and key limits

Free allows 2 KiB and 10 keys; Pro allows 8 KiB and 30 keys; Scale allows 32 KiB and 100 keys. An absolute 32 KiB metadata ceiling also applies.

## Rate limits

Each API key has its own per-minute window:

| Account plan | Requests per minute per key |
| ------------ | --------------------------: |
| Free         |                          60 |
| Pro          |                         600 |
| Scale        |                       3,000 |

A full window returns `429 rate_limited`, a `Retry-After` header, and `retryAfterSeconds` in the error details. One busy key does not consume another key's window.

## Monthly event capacity

The workspace account plan also limits monthly runtime events. Idempotent replays do not claim a second event slot.

## Credit concurrency

Concurrent consumes cannot overdraw a balance. If twenty calls compete for a balance of ten credits at a ten-credit cost, exactly one succeeds and the other nineteen are denied with `insufficient_credits` — no partial charges, no negative balance.

## Idempotency concurrency

Concurrent requests with the same key wait for the original owner to finish and then receive the recorded response. A shared key produces one logical charge, one event, and one ledger row.

## Project isolation

The API key resolves the project for every request. A customer ID, plan, feature, or idempotency key from another project cannot be selected by changing the request body.