A structured lesson workspace with readable content, hands-on examples, and a clean path to completion.
Monitoring tells you what happened (the workflow failed). Observability tells you why it happened (the LLM passed a string instead of an integer). Evaluation tells you if the outcome was actually correct.
Agent monitoring and observability
This video requires an access code. Enter your code below to unlock it instantly.
Don't have a code? Contact us to get access.
Open the n8n execution logs. Look at the payload the agent sent to the calculator tool. A common failure occurs when the user types '100 square feet'. The LLM passes the literal string '100 square feet' into the size_sqft parameter instead of the integer 100. The JavaScript node throws a type error.
{"error": "TypeError: Cannot perform arithmetic on string '100 square feet'","tool": "calculator_tool","inputReceived": "100 square feet"}
Without observability, the agent just replies 'I encountered an error.' With observability, you see the exact type mismatch and can add a regex sanitization step in the sub-workflow to strip the text before the math runs.
Use the n8n Evaluations tab to run test datasets against your agent. If you are not running automated evals, you are guessing your accuracy.
Finish the lesson once you have worked through the material. This awards ★ 30 XP.