How to resolve 429 rate limit errors when sending prompts to Claude from n8n?

The short version:
To fix 429 errors when sending prompts to Claude from n8n, you need to slow down and spread out your requests. In n8n, this usually means adding queueing, a Rate Limit node or Wait nodes, retries with backoff, and avoiding parallel execution. Claude rate limits are strict, and if your workflow sends too many messages too quickly, Anthropic will return 429. The reliable fix is to control throughput inside n8n so you never exceed what the API allows.

 

Why 429 Happens

 

Claude responds with HTTP 429 when you're sending more requests per minute than your API plan allows. This isn't an n8n bug — it's Anthropic protecting their API from overload. n8n just forwards the error.

Because n8n runs nodes as fast as possible unless you slow them down, your workflows can unintentionally hit rate limits, especially when iterating over items or when multiple executions run at once.

 

Practical Fixes That Work in Production

 

Here are the things that actually work in real n8n deployments when dealing with Claude rate limits:

• Force sequential requests so you’re not hitting the API in parallel.
• Add a Wait or Rate Limit node between messages to Claude.
• Add retry with exponential backoff to handle occasional 429 bursts.
• Use a queue pattern (Webhook → Save to DB → Scheduled Poll → Send to Claude) if you have high volume.
• Reduce per-prompt size to avoid longer processing time, which can indirectly cause bursts.

 

Step‑by‑Step: How to Make Your Workflow Stop Hitting 429

 

Below is a practical approach that is known to work reliably.

• Put a Rate Limit node before every Claude API call.
  Configure it to something safe like:
  Limit: 1 request, Per: 1 second (adjust based on your Anthropic plan).
• Make sure the Claude node processes items one by one.
  In n8n, go to the Claude node → Settings → set:
  Run Once for All Items: true
  This prevents it from blasting many parallel API calls.
• Add retry logic directly in the Claude node.
  Settings → Retry On Fail → true
  Settings → Max Attempts: 5
  Settings → Retry Delay: 5000 ms (or use 10000 ms for heavy workloads)
• For bulk operations, split into batches.
  Use the Item Lists or Split In Batches node and process in controlled chunks.
• If you have multiple workflows using Claude, unify them behind a queue.
  The reliable pattern:
  Webhook or Trigger → Write to DB → Scheduled workflow reads next job → Calls Claude → Saves result
  This ensures only one request is sent at a time.

 

Example: Using a Function Node for Exponential Backoff

 

This is useful when you want finer control and don’t want to rely only on the node’s built‑in retry.

// Simple exponential backoff value generator
// Use this expression for a Wait node afterwards

const attempt = $json.attempt || 1     // your workflow should increment this
const baseDelay = 3000                 // 3 seconds
const delay = baseDelay * Math.pow(2, attempt - 1)

return [{ delay }]

You then feed {{ $json.delay }} into a Wait node’s "Wait Amount" field.

 

Extra Tips That Matter in Production

 

• Check your Anthropic dashboard to confirm your rate limits. Different plans have different caps.
• Make sure no other workflows hit Claude simultaneously.
• Pay attention to item loops. A loop of 50 items can create 50 API calls instantly unless rate‑limited.
• Use n8n Cloud’s concurrency controls or, in self‑hosted setups, reduce worker concurrency if needed.

 

The Real Core Idea

 

The only reliable way to eliminate 429 errors in n8n when calling Claude is to limit throughput. Add a Rate Limit node, run nodes sequentially, use retries with backoff, and centralize requests if you have high volume. Once you slow your traffic to match Anthropic’s limits, the errors disappear.