Help
How to read an agent run timeline
An AI that answers your customers is only acceptable if you can find out afterwards what it did. The run timeline is that record.
3 min read
What is a run?
One invocation of the agent: a customer message came in, the agent read its instructions, called some tools, and produced an answer.
Each run is stored against the agent, so you can look back through what it has been doing rather than only watching it live.
What does the timeline show?
Each step with its status, which tools were called, the input each one received and what it returned.
That is enough to distinguish the two different failures. If the agent called the right tool and got the right data but answered badly, the instructions need work. If it never called the tool, a permission or an instruction is missing.
How do you use it to improve answers?
Find the step where the run diverged from what you wanted, then change the instruction or the data that caused it.
Recurring wrong answers usually trace to missing product data rather than a misbehaving model. If customers keep asking something your product pages do not state, no instruction will make the agent know it.
Common questions
Are timelines kept forever?
Execution history is retained according to the app's data retention policy, not indefinitely. Do not treat it as an audit archive.
Can I see what the agent said to a specific customer?
Yes. The conversation history and the run that produced each reply are both available.
Related guides
- How to set up an AI agent that answers from your live inventory
- How to hand a WhatsApp conversation to a human
All guides are listed in the help centre. If none of them answers it, email [email protected].
Start with the free plan
Free: 1 agent, 100 messages a month and up to 100 products indexed. No credit card required, billed through Shopify, cancel any time.