How can I prevent simultaneous HTTP requests from causing a 429 error in the AI Agent node?

Below is an explanation and some workarounds to manage your rate limit (429) issue when using the AI Agent node:

---

Why the 429 Error Occurs

The error “Request failed with status code 429” means that too many requests are being sent to the API in a short amount of time. The AI Agent node triggers these requests in parallel, which can exceed the API’s rate limit.

---

Workarounds

Since the AI Agent node does not offer built‑in options like batching or “retry on fail” (features available in the HTTP Request node), you can try one of these strategies:

1. Throttle Your Workflow Manually
   Use flow control nodes such as Loop Over Items and Wait to split your data into smaller chunks and introduce a delay between batches. This way, you avoid sending all requests simultaneously.

   For example, insert a Loop Over Items node before the step that triggers the API call, then after the API call, add a Wait node to create a delay before looping to the next set of items.

2. Switch to the HTTP Request Node
   If possible, consider using the HTTP Request node instead of the AI Agent node. The HTTP Request node has built‑in options for:

   • Batching: You can bundle items and set a delay between each batch.
   • Retry on Fail: Automatically retry the API request after waiting a set amount of time.

   For example:

   JSONCopied
   {
       "nodes": [
           {
               "parameters": {},
               "name": "Manual Trigger",
               "type": "n8n-nodes-base.manualTrigger",
               "typeVersion": 1,
               "position": [300, 300]
           },
           {
               "parameters": {
                   "batchSize": 100,
                   "options": {}
               },
               "name": "Batch Items",
               "type": "n8n-nodes-base.splitInBatches",
               "typeVersion": 3,
               "position": [500, 300]
           },
           {
               "parameters": {
                   "url": "https://api.example.com/endpoint",
                   "options": {},
                   "requestOptions": {},
                   "retryOnFail": true,
                   "maxTries": 3,
                   "waitBetweenTries": 1000
               },
               "name": "API Request",
               "type": "n8n-nodes-base.httprequest",
               "typeVersion": 1,
               "position": [700, 300]
           },
           {
               "parameters": {
                   "unit": "seconds",
                   "time": 1
               },
               "name": "Wait",
               "type": "n8n-nodes-base.wait",
               "typeVersion": 1,
               "position": [900, 300]
           }
       ]
   }

   This example shows how to split the items into batches, make the API request with retry logic, and wait between batches to remain within rate limits.

---

Summary

• The AI Agent node sends requests in parallel, which can hit API rate limits.
• Since it lacks built‑in batching or retry on error, consider throttling your workflow manually by combining Loop Over Items and Wait nodes.
• Alternatively, switch to using the HTTP Request node, which includes built‑in support for batching and retrying on request failures.

These approaches should help you manage the rate of outgoing requests and avoid the 429 error in your deep search workflow.