Issue in the N8N AI Assistance

I am using the trail version of the N8N to see how well it suits me. After generating some workflows using N8N AI assistance, whenever I send any message to N8N AI Assistance, it throws an error message asking me to try after some time. Despite waiting and trying with another account and another system, I could not resolve the issue. Could you please help me with this?

Describe the problem/error/question

Error: Something went wrong before I could finish that response. Please try again.
Something went wrong before I could finish that response. Please try again.

What is the error message (if any)?

Please share your workflow



Share the output returned by the last node

Information on your n8n setup

  • n8n version:
  • Database (default: SQLite):
  • n8n EXECUTIONS_PROCESS setting (default: own, main):
  • Running n8n via (Docker, npm, n8n cloud, desktop app):
  • Operating system:

Hi @DRDO_Data_Collection

Did you run out of credits?

Thanks for letting us know about this, We have created CV-7 as the internal dev ticket to look into it.

@kjooleng I don’t think so. As shown in the screenshot, I have some free credits.

Oopss…my bad…

Hi @DRDO_Data_Collection Welcome!
On Cloud, that message is what the AI Assistant returns when its AI credit pool is exhausted, the wording just doesn’t say so:

The credits in your screenshot are the workflow generation credits, a separate pool from the Assistant’s own AI credits. To confirm, open the browser network tab and send one message to the assistant: if GET /rest/instance-ai/credits returns 500 while the chat, threads, and preferences calls return 200, that pool is empty and the account needs more AI credits before the assistant responds again.

Hi @Anshul_Namdev ,

Many thanks for your response. Though your response is clear. Still I am not convience, as I don’t see any error message in the network tab (refer to the screenshot). If you allow me, I want to ask you the following question for better clarity as it is not clear from the website:

  1. one AI credit is equal to how many tokens and how the N8N measures the workflow generated through AI assittance in terms of tokens?

Hi @DRDO_Data_Collection
n8n doesn’t meter the AI Workflow Builder in tokens, so there’s no fixed “one credit = X tokens” figure. Each message you send to the builder to create or modify a workflow counts as one interaction, and one interaction is one credit, regardless of how many tokens that turn uses behind the scenes. A short prompt and a long, complex one both cost a single credit.

Plan by plan credit allowances are here:

@Anshul_Namdev Thank you for the clarity.

Could you please confirm that the error I am getting is becuase of the AI credits, as I am not seeing the 500 error in the network tab?

If so, the N8N says free trail has 800 AI tokens. But I hardly interacted 20 to 30 times with the AI assistance.

Hi @DRDO_Data_Collection
No, I can’t confirm it’s your credits, and the signs point the other way. You said a second account hit the same error, and a separate account gets its own credit allowance, so if a lack of credits were the cause, that fresh account would have worked. The credits call not returning a 500 says the same thing: the credit check is passing, so exhaustion isn’t what’s blocking the assistant. This looks like a server-side problem, not your balance.
A failure on n8n’s side won’t always show as a failed request in your own network tab, so the confirmation has to come from them. Email help@n8n.io with your Cloud instance URL and the timestamp of a failed attempt so they can match it against the server logs.

@Anshul_Namdev Roger that. Many thanks!

@DRDO_Data_Collection if any of my reply have really helped you can consider marking that as solution or may be in future. Thanks!

We have taken a look at this issue and are unable to confirm that this is a bug, For now we have closed the internal ticket but if it starts to look like it is a bug our moderation team will flag this again.

Hello, I have exactly the same problem. The AI assistant is unusable and I’m getting the same error message.