A8gent
GlossaryAutomation mechanics

What is rate limit?

Also called: API rate limiting, throttling, 429

A rate limit is the cap a service puts on how many requests you may make in a period, enforced by rejecting excess calls - usually with HTTP status 429 - to protect the service from overload.

Written by A8gent
Contents
  1. At a glance
  2. What a rate limit is
  3. How limits are actually measured
  4. Designing around them
  5. Cost and the hidden interactions
  6. What you should know
  7. FAQ
  8. Related terms

At a glance

Category
Automation mechanics
Status code
429 Too Many Requests (RFC 6585)
Measured by
Requests per interval, concurrency, tokens per minute, or query cost
First thing to check
Response headers for remaining quota and reset time, and Retry-After on a 429
AI providers
Usually cap requests per minute and tokens per minute separately
Design tools
Batching, queueing, exponential backoff with jitter, webhooks instead of polling

What a rate limit is

A rate limit is a service's cap on how much you may call it. Exceed it and requests are rejected, conventionally with HTTP status 429 Too Many Requests, defined in RFC 6585. Every serious API has limits, and hitting them is not an error in your code so much as a fact of the environment your automation lives in.

They exist for good reasons - one customer's runaway loop should not degrade the service for everyone - and they are the reason a workflow that worked with ten records fails at a thousand. That failure mode is the single most common surprise when an automation moves from testing to real volume.

How limits are actually measured

More ways than people expect. Requests per interval is the classic: N per second, per minute, per day. Concurrency caps how many requests may be in flight at once, regardless of rate. Token-based limits apply to AI providers, which typically cap both requests per minute and tokens per minute, so a few large prompts can hit the limit before the request count is anywhere near it.

Cost-based limits appear in GraphQL APIs, where each query is scored by complexity and you spend a budget rather than a count. And limits often differ by endpoint, by plan tier, and between sandbox and production. Read the specific numbers rather than assuming, and check whether the limit is per key, per account or per IP - it determines whether separate keys actually help.

Designing around them

Five techniques, roughly in order of value. Read the response headers: most APIs return how many requests remain and when the window resets, so you can slow down before being rejected rather than after. Respect Retry-After when a 429 includes it - it is the server telling you exactly how long to wait.

Back off exponentially with jitter on any 429 or 5xx, which is covered under retry and backoff. Batch wherever a bulk endpoint exists, since one call for a hundred records beats a hundred calls. And queue work rather than firing it all at once, so a burst is smoothed into a steady rate - see queue.

Cost and the hidden interactions

Rate limits interact badly with two other things. Polling consumes your limit continuously whether or not there is work, which is another argument for webhooks. And retries consume it too, so a naive retry loop during an outage can turn a brief problem into a sustained lockout - the reason backoff exists rather than a fixed retry interval.

There is also a billing interaction worth naming: on per-action automation platforms, a rejected call may still consume your billing unit. Hammering a rate-limited API therefore costs money while achieving nothing.

What you should know

Find the limits before you build, not after. They belong in the same pre-project checklist as authentication method and webhook support, because a limit of sixty requests per minute against a nightly sync of fifty thousand records is a design constraint, not a detail.

Then make hitting the limit visible. A 429 that is silently retried forever looks identical to a working system until someone notices records are hours behind. Log them, count them, and alert when the rate rises - in observability terms, rate-limit pressure is a leading indicator that a workflow has outgrown its design.

Rate limit FAQ

What does HTTP 429 mean?

Too Many Requests - you have exceeded the service's rate limit and the request was rejected rather than processed. It is a transient condition, so it should be retried after a wait, ideally the wait specified in the response's Retry-After header.

How do I find an API's rate limits?

The documentation states them, and most APIs also return them in response headers showing remaining quota and reset time. Check whether limits are per key, per account or per IP, and whether they differ by endpoint and plan tier - those details change your design.

How do rate limits work for AI providers?

Usually two limits at once: requests per minute and tokens per minute. Large prompts can exhaust the token limit while the request count is low, which is why an agent with a big context can be throttled even though it is making few calls.

What should my automation do when it hits a rate limit?

Wait and retry with exponential backoff plus jitter, honouring any Retry-After header. Do not retry immediately in a tight loop - that extends the lockout, burns your platform's billing unit on rejected calls, and can escalate a brief throttle into a sustained one.

How do I avoid hitting rate limits at all?

Batch requests where a bulk endpoint exists, queue work so bursts are smoothed into a steady rate, replace polling with webhooks so you are not calling constantly, and read the remaining-quota headers so you can slow down before being rejected rather than after.

Do rejected calls still cost me money?

On per-action and per-operation automation platforms, often yes - the step ran, it just failed. So a workflow stuck in a retry loop against a rate-limited API burns billing units continuously while accomplishing nothing, which is why backoff and alerting both matter.

Primary source

RFC 6585 - Additional HTTP Status Codes (429)

Checked August 2026. Anything priced or rate-limited moves, so verify against the source before you budget.

Want this built rather than explained?

Knowing what rate limit means is one thing; having it running in your business is another. We scope one workflow at a time, quote a fixed price in writing after a free consult, build it, and hand it over with documentation you own.