A8gent
All articles

Technical

The service is receiving too many requests from you: 6 Causes and the Fix for Each in n8n

This is the message n8n shows when the service you are calling answers with HTTP 429. n8n is not the one limiting you. The fix depends on whether you are sending too fast, sending too much in total, or out of quota, and each of those has a different test and a different node setting.

A8gentAutomation practiceNovember 17, 202614 min read
Contents
  1. The Short Answer
  2. What the Error Actually Means
  3. Cause 1: One Node Is Sending Every Item at Once
  4. Cause 2: A Per-Minute Quota on an App Node
  5. Cause 3: Your LLM Provider's Rate Limit
  6. Cause 4: It Is Not Speed, It Is Quota or Billing
  7. Cause 5: Retry On Fail Is On, but the Window Is Longer Than the Retries
  8. Cause 6: Several Executions Share One Limit
  9. How to Stop It Coming Back
  10. Sources
  11. FAQ

The Short Answer

"The service is receiving too many requests from you" is the text n8n attaches to any response with HTTP status 429. The limit belongs to the service the node is calling. n8n is reporting it, not enforcing it.

Three fixes cover almost every case:

  1. HTTP Request node: select Add Option > Batching, set Items per Batch to 1 and Batch Interval (ms) to 1000 for an API that allows one request per second.
  2. Any node, short limits: open the node's Settings, enable Retry On Fail, and set Wait Between Tries (ms). The ceiling is 5000 ms and five tries.
  3. Any node, longer limits: put a Loop Over Items node before the node that calls the API and a Wait node after it, connected back to Loop Over Items.

Before you change anything, read the second line of the error. n8n shows the service's own message under its label, and that message tells you whether you were too fast, over a daily allowance, or out of credit. Those need different fixes, and the rest of this page is one section per cause with a test for each.

What the Error Actually Means

n8n keeps a small table of friendly messages for common HTTP status codes. When a node gets an error response, n8n looks up the status code and shows the matching text. The entry for 429 is exactly The service is receiving too many requests from you. You can read the table in the n8n source, in node-api.error.ts, next to the ones you have probably also met:

StatusMessage n8n shows
400Bad request - please check your parameters
401Authorization failed - please check your credentials
403Forbidden - perhaps check your credentials?
404The resource you are requesting could not be found
429The service is receiving too many requests from you
503Service unavailable - try again later or consider setting this node to retry automatically (in the node settings)

Older n8n versions used a longer wording, "The service is receiving too many requests from you! Perhaps take a break?", which is why you will see both in forum threads. They are the same error.

The n8n rate limit documentation puts it plainly: when a node hits a rate limit it errors, n8n displays the error in the node output panel, and that includes the error message from the service. So the useful part is not the headline. It is the detail under it, for example Google's Quota exceeded for quota metric 'Read requests' and limit 'Read requests per minute per user'. That line names the limit. Find it before you guess.

Cause 1: One Node Is Sending Every Item at Once

This is the common one. n8n nodes process a list of items, and for an API call that means one request per item. The n8n docs say it directly for the HTTP Request node: each input item generates a separate HTTP request. Feed the node 300 rows and it sends 300 requests back to back.

How to confirm it: open the failed execution and look at the item count going into the failing node. If it is more than a handful and the workflow runs fine when you test it with one or two items, you have a burst.

The fix on the HTTP Request node is the built-in Batching option:

  1. In the HTTP Request node, select Add Option > Batching.
  2. Set Items per Batch to the number of input items to process in each batch.
  3. Set Batch Interval (ms) to the delay between batches. 1000 waits one second.

Work the numbers back from the API's published limit. If the service allows 60 requests per minute, one item per batch with a 1000 ms interval keeps you at the limit, and 1200 ms gives you headroom. If it allows 10 per second, ten items per batch with a 1000 ms interval does the same job faster.

Batching only exists on the HTTP Request node. For app nodes (Google Sheets, Slack, HubSpot and so on) use the loop pattern in the next section.

Cause 2: A Per-Minute Quota on an App Node

App nodes have no Batching option, and many of the services behind them count requests per minute per user. Google Sheets is the one people hit most. In a long-running n8n community thread the detail under the error read Quota exceeded for quota metric 'Read requests' and limit 'Read requests per minute per user', and the fix that worked was a Wait node inside a loop.

How to confirm it: the service's message mentions a per-minute or per-user quota, and the node fails part of the way through a large set of items rather than on the first one.

The fix is the pattern the n8n docs recommend for integrations:

  1. Add a Loop Over Items node before the node that calls the API. Set Batch Size to the number of items to release on each pass.
  2. Connect the loop output to your API node.
  3. Add a Wait node after the API node. Set Resume to After Time Interval, then set Wait Amount and Wait Unit.
  4. Connect the Wait node back to Loop Over Items. The loop's done output carries on to the rest of the workflow once every item has been processed.

The Wait node only throttles anything when it sits inside the loop. A Wait placed after a node that has already processed every item does nothing for the rate, because all the requests have gone out by then. The same thread makes the point: the Wait trick works when the records are being processed through the batching loop.

Look for a cheaper fix first, though. A second poster in that thread was hitting a write quota because a Clear operation ran once for every incoming item, around 65 times. Moving it so it ran once at the start removed the problem without any waiting. If a node only needs to run once, open its Settings and check whether it is being executed per item when it should not be. Fewer calls beats slower calls.

Cause 3: Your LLM Provider's Rate Limit

The OpenAI node, chat model sub-nodes and AI Agent workflows surface the same message, because the model provider answers with 429 like any other API. The OpenAI node common issues page lists this exact error and says it displays when you have exceeded OpenAI's rate limits.

How to confirm it: the failing node is a model call, and the provider's message refers to requests per minute or tokens per minute. Your provider's dashboard shows the limits for your account tier.

The fix from the n8n docs is one of two things:

  • Split the data into smaller chunks with Loop Over Items and add a Wait node at the end of the loop, as in Cause 2. The example workflow n8n publishes for this uses a batch size of 500 and a Wait measured in minutes, which gives you a sense of the scale intended: big batches, long pauses.
  • Call the provider's API with the HTTP Request node instead of the OpenAI node, so you can use the Batching option from Cause 1.

Token-per-minute limits behave differently from request limits. Ten requests carrying long documents can exceed a token allowance that a hundred short ones would not. If your prompts are large, shrink what you send before you slow down how often you send it. Our guide to reducing AI agent LLM costs covers trimming context, and the same changes lower your token rate.

Cause 4: It Is Not Speed, It Is Quota or Billing

A 429 does not always mean "slow down". Some services use it when an account has run out of allowance for the day, the month, or the plan. Slowing the workflow down will not help, and retrying makes the logs noisier without changing the outcome.

How to confirm it: a single manual run with one item fails straight away. A rate limit caused by speed does not fail on the first request after a quiet period. Then read the service's message for words like quota, billing, credit, or plan.

The fix is on the service's side, not in n8n. For OpenAI specifically, the n8n docs have a separate entry for Insufficient quota with a checklist worth following in order:

  • Check that the organization has credit on the billing overview page, and that you are looking at the correct organization and project for the API key in your n8n credential.
  • Check the organization and project usage limits.
  • If you have just topped up, expect a delay before the balance is reflected. The docs also suggest issuing a new API key after adding credit.
  • In n8n, confirm the node is using the credential for the account you added money to.

The same logic applies to any provider with a free tier. If the free allowance is gone, the workflow needs a paid plan or a different model, not a Wait node.

Cause 5: Retry On Fail Is On, but the Window Is Longer Than the Retries

Retry On Fail looks like the obvious answer and it is the right one for brief spikes. It is also the reason some workflows still fail after "adding retries". The n8n docs state the limits:

  • Wait Between Tries is limited to 5000 ms.
  • Max Tries is limited to five.
  • The HTTP Request node does not automatically honor the API's Retry-After response header.

Five tries at five seconds is roughly twenty seconds of patience. If the service resets its counter every minute, or tells you to come back in 60 seconds, every retry lands inside the same closed window and the node fails anyway.

How to confirm it: the execution shows the node attempted several times before failing, and the service's limit is per minute or longer.

The fix: keep Retry On Fail for what it is good at, and add pacing for the rest.

ToolWhere it livesGood forLimit
Retry On FailNode > SettingsOccasional 429s and per-second limitsFive tries, 5000 ms between tries
BatchingHTTP Request > Add OptionSteady pacing of many itemsHTTP Request node only
Loop Over Items + WaitTwo extra nodes on the canvasPer-minute and longer limits, any nodeMore nodes to maintain

Retry On Fail reacts after the limit has been hit. Batching and the loop stop you hitting it. For a workflow that runs on a schedule with predictable volume, prevention is the better default, with a small retry on top for the odd blip. The general pattern is described in our glossary entry on retry and backoff.

Cause 6: Several Executions Share One Limit

You paced one workflow carefully and it still gets 429s. The usual reason is that the service counts requests per API key or per user, and more than one execution is using that key at the same moment. A webhook-triggered workflow is the classic case: ten webhook calls arrive together, n8n starts ten executions, and each one sends its "one request per second" at the same time.

The n8n docs on concurrency control confirm the behaviour: in regular mode, self-hosted n8n does not limit how many production executions run at the same time unless you tell it to.

How to confirm it: open the Executions list and look at start times around the failure. Overlapping executions of the same workflow, or of other workflows using the same credential, point here.

The fix on self-hosted n8n is a production concurrency limit:

export N8N_CONCURRENCY_PRODUCTION_LIMIT=5

Executions over the limit are queued and processed in order as capacity frees up. Two details from the docs matter: the limit applies only to production executions started from a webhook or trigger node, not to manual runs or sub-workflow executions, and it is disabled by default. On n8n Cloud, concurrency limits are set by plan and shown at the top of the executions tab.

If you cannot change instance settings, the workflow-level alternative is to stop doing the API work inside the webhook workflow. Have the webhook write the job to a table or a queue, and let one scheduled workflow drain it at a controlled pace with the loop from Cause 2.

How to Stop It Coming Back

Once the immediate failure is fixed, four habits keep it fixed.

1. Look up the limit and write it on the canvas. The n8n docs say to check the API documentation of the service for its limits. Do that once, then leave a sticky note next to the node with the number and the date you checked. The person who doubles the input volume in six months will need it.

2. Pace by default on anything that loops. If a node will ever receive more than a few items, add Batching or the loop now. It costs a few seconds per run and removes a class of 3 a.m. failures.

3. Fetch less. If you are reading a large dataset page by page, use the HTTP Request node's pagination settings rather than firing many separate requests, and filter on the service's side so you do not request records you will throw away.

4. Make failures loud. Set an error workflow in Workflow Settings so a failed execution sends you a message. The n8n docs describe the setup: a separate workflow that starts with the Error Trigger node, selected under Error workflow in the settings of each workflow you want covered. A 429 you hear about in a minute is a small problem. One you find out about from a customer is not. See error handling for the wider pattern.

If you are building workflows that call models and APIs in volume, the pacing, retry and alerting patterns here are a core part of our n8n AI Agents course. For how n8n compares with other platforms on execution limits and pricing, see our n8n platform review, and to estimate what a paced workflow costs to run, use the n8n workflow cost calculator.

Sources

Everything on this page was checked against n8n's own documentation, source code and community forum on 17 November 2026.

FAQ

What does "The service is receiving too many requests from you" mean in n8n?
It is the fixed message n8n shows when the service a node called responded with HTTP status 429, too many requests. The limit is set by that service, not by n8n. The service's own explanation is shown under the message in the node output panel and usually names the exact limit, such as requests per minute per user.
Is n8n rate limiting me?
No. n8n is passing on a 429 response from the API or app you are calling. n8n Cloud does limit concurrent production executions by plan, but executions over that limit are queued rather than failed, and that is a different mechanism from this error.
What should I set Batch Interval to in the HTTP Request node?
Work it out from the API's documented limit. For an API that allows one request per second, set Items per Batch to 1 and Batch Interval (ms) to 1000. For 60 requests per minute the same setting sits exactly on the limit, so use a slightly larger interval for headroom. The option is under Add Option > Batching.
Why does Retry On Fail not fix my 429 errors?
Retry On Fail is capped at five tries with at most 5000 ms between them, and the HTTP Request node does not read the Retry-After header. If the service's limit resets every minute or longer, all five retries land inside the same blocked window. Use the HTTP Request Batching option or a Loop Over Items node with a Wait node for longer limits.
How do I slow down a Google Sheets, Slack or other app node that has no Batching option?
Add a Loop Over Items node before it with a small Batch Size, add a Wait node after it set to resume After Time Interval, and connect the Wait node back to Loop Over Items. The Wait node only throttles the requests when it sits inside the loop.
I get the error on the very first request. Is that still a rate limit?
Probably not a speed limit. A failure on a single request after a quiet period usually means the account is out of quota or credit, or is on a tier that does not include the endpoint. Read the service's message under the n8n error and check the account's billing and usage pages. For OpenAI, the n8n docs list this separately as Insufficient quota.
Does this error mean my workflow lost data?
The execution stopped at the failing node, so items after the point of failure were not processed in that run. Items already processed before the failure were sent. Before re-running, check whether the earlier calls had side effects such as created records, so that a retry does not duplicate them.

If this resonated

Module 02 is the long version of this post.

Eval set, refusal pattern, kill switch. Build it in a weekend.

All postsLast verified 2026-11-17

Done for you

Want this built for your business?

Tell us the workflow. We scope, build, and ship the agent with guardrails and the numbers to prove it worked. The scoping call is free.

Free scoping call. You own the code.

Request a scoping call