Ai node problems

Questions
1

My AI Agent node keeps returning incomplete responses. What should I check?

Describe the problem/error/question

What is the error message (if any)?

Please share your workflow

(Select the nodes on your canvas and use the keyboard shortcuts CMD+C/CTRL+C and CMD+V/CTRL+V to copy and paste the 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:
2

Some models have limited output tokens. What model did you use, @Korede_Ojeyinka ?

3

Start by reviewing the model configuration, token limits, and prompt structure. Incomplete responses often occur when the output exceeds the maximum token limit or when the prompt lacks clear instructions. Monitoring execution logs and testing with shorter prompts can help identify the root cause.

4

Can you share what model you’re using, the exact agent type (Tools Agent / Conversational Agent), and whether the responses are cut off mid-sentence or just missing parts of the expected output? Also check the AI node’s output in the execution log - if the text field looks complete there but something is missing downstream, the issue is likely in how a later node reads the output rather than the model itself.

5

Hey @Korede_Ojeyinka, to help narrow it down while you share more details, “incomplete” usually means one of three things:
Cut off mid-sentence → max_tokens is too low. Increase it in the model node settings e.g. from 1000 to 4000+.
Missing expected sections → your system prompt isn’t specific enough. Tell the model explicitly what structure you expect e.g. “Always return sections: Summary, Steps, and Conclusion”.
Output looks complete in the execution log but missing downstream → you’re likely referencing the wrong field. Check that you’re using {{ $json.output }} and not an expression pointing at a field that doesn’t exist.
Sharing your system prompt and a sample of what the node actually outputs from the execution log will make it much easier to pinpoint which one it is.