How can I resolve the 'Binary data is not supported' error when using the HTTP Request Tool in an AI Agent workflow?

Below is an explanation of why you’re seeing the error and some steps to resolve it.

---

Why the Error Occurs

When using the standard HTTP Request node, you have full control over how the response is processed. In contrast, the HTTP Request Tool node is designed specifically for an AI agent context and automatically “optimizes” the response for text format. If the API call returns data that is evaluated as binary (for example, if the Content-Type isn’t explicitly text‑based or if the API returns a file-like response), the node will throw an error stating that “Binary data is not supported.”

Even though you have set "responseType": "text" and enabled "optimizeResponse": true, the configuration may still lead the tool to treat the response as binary if the ERP system’s endpoint does not return the expected text or JSON MIME type. One potential cause is that appropriate headers (such as an "Accept" header to request text or JSON) might be missing or misconfigured in the HTTP Request Tool setup.

---

How to Resolve the Issue

1. Ensure the API Returns a Text-Based Response:

   • Double-check the ERP system’s API documentation to see which MIME type is returned for that endpoint. If it’s returning something like application/octet-stream (a generic binary type), you may need a change on the API side or an additional parameter on your side to request a text-based format.
   • If possible, adjust your API call to include an appropriate Accept header. For example, to request a JSON response add:

     JSONCopied
     {
       "Accept": "application/json"
     }

   • In the HTTP Request Tool node, make sure you set the header under the JSON headers section or via a header parameter.

2. Disable or Reconfigure the Response Optimization:

   • If adjusting the API isn’t feasible, try disabling "optimizeResponse". This will have the node return the full response without attempting to convert it; you can then process it in later nodes.
   • Alternatively, verify that the responseType setting remains set to "text" so the node knows to expect textual data.

3. Use a Standard HTTP Request Node If Needed:

   • If the ERP endpoint is designed to return binary data (such as a file or a PDF), then the HTTP Request Tool (meant for integrations with AI agents expecting text) isn’t appropriate. In that case, continue using the standard HTTP Request node or add a conversion step (for example, using a Function node to convert the binary data into a text string, if that makes sense for your application).

---

Example – Forcing a JSON Response via Headers

Here’s an example where you explicitly set the Accept header:

JSONCopied
{
  "parameters": {
    "toolDescription": "Rufe den Auftrag 63144 auf.",
    "url": "https://637a04a74b2c1.xentral.biz/api/salesOrders/63144",
    "sendHeaders": true,
    "specifyHeaders": "json",
    "jsonHeaders": "{\"Accept\": \"application/json\"}",
    "optimizeResponse": true,
    "responseType": "text"
  },
  "type": "@n8n/n8n-nodes-langchain.toolHttpRequest",
  "typeVersion": 1.1,
  "name": "Auftrag 63144 abrufen"
}

By ensuring the request explicitly asks for JSON, you help n8n to treat the response as text.

---

Summary

The HTTP Request Tool node in the AI Agent context is optimized for text responses. If your API returns data that is evaluated as binary, you’ll see an error. To resolve this, confirm the ERP endpoint’s response type, explicitly request a text‑based (or JSON) response with the proper headers, adjust the optimize settings, or use the standard node if binary data is required.

Let me know if you have any other questions about configuring HTTP requests in n8n!