# Rate limits

> Each live key may send a number of requests a second, set by its plan. Over it, the API answers `429 RATE_LIMITED` with `Retry-After`. Sandbox keys have no rate limit.

Base URL: `https://api.publicadsapi.com` · Auth: `authorization: Bearer <key>` · Web version: https://publicadsapi.com/docs/rate-limits

## Requests a second, per live key

| Plan | Requests a second |
| --- | --- |
| Free | 5 |
| Starter | 20 |
| Growth | 50 |
| Enterprise | Set for your traffic |

Counted over the last second, per key. Over it, `429 RATE_LIMITED` with `Retry-After` in seconds; nothing is charged. Since a live call takes seconds, a key rarely meets this limit: what you will meet first is how many calls you run at once.

## Calls at once

Live calls run in our browsers, and there are only so many. When every one is busy the API answers `503 BUSY` with `Retry-After` (a few seconds) rather than queueing, and nothing is charged. If you need many calls running at once, [talk to us](https://publicadsapi.com/contact).

## The platforms' own limits

The libraries limit traffic too. We spread calls over many connections and retry blocked ones, so you don't have to. When a platform blocks every attempt the answer is `503 UPSTREAM_BLOCKED` with `Retry-After`, never charged. LinkedIn blocks most often.

## Unclaimed agent accounts

- Agent signup: 20 accounts a day from one network, with a cap across all agents. Over it, `429 RATE_LIMITED` with `Retry-After` until midnight UTC.
- Live calls from unclaimed agent accounts: 20 a day from one network, with a cap across all agents, besides the 50 free credits a month. Over it, `429 RATE_LIMITED`.
- Once a person claims the account, the plan's limits apply instead.

`X-RateLimit-*` headers on live calls count this month's included credits, not requests a second: see [Credits and billing](https://publicadsapi.com/docs/credits.md#headers).

See also: [Credits and billing](https://publicadsapi.com/docs/credits.md) · [Agents](https://publicadsapi.com/docs/agents.md)