Action Steps

In the Playground and Logs, see the sources and actions behind each reply: what ran, what it sent and received, and why other actions didn't run.

Under each agent reply in the Playground, a line shows how the reply was made: the knowledge sources it read and the actions behind it. Select it to see every step.

In Activity > Logs, conversations show the same summary line and action steps, plus the widget the visitor saw. The widget is read-only. Visitors never see action steps.

Steps appear live. While an action runs, the line reads Running Check order status…, then it changes to the result before the reply arrives. Visitors see only the action’s While it runs message, such as “Looking up your order…”. See Custom API.

The summary line

Summary	Meaning
Read 2 sources	The agent used two knowledge sources for its answer. Expand it to see their links.
Used Check order status	The action ran. Shown in red with 1 failed if it failed.
Used Check order status · 3 didn’t run	One action ran, and three others were checked but didn’t match.
Send an idea is waiting for details	The action matched, and the agent is asking the visitor for details it needs first.
No actions ran	Actions were checked, but none matched this message.

Step statuses

Status	Meaning
Running	The action is running now. It changes to Succeeded or Failed when it finishes.
Succeeded	The action ran. Sent shows the values it sent, and Received shows what it got back, which is what the agent used.
Failed	The action ran but failed, for example because your API returned an error. Received shows the error. The agent tells the visitor it couldn’t be completed.
Waiting for details	The message matched When to use, but a required value was missing. The agent asked for it, and the visitor’s answer completes the action.
Matched	The message matched When to use, and the action continued in its own way, for example by showing a lead form.
Didn’t match	The message didn’t match When to use, so the action didn’t run. The step says Almost matched when the message came close.
Already ran	The action already ran in this conversation, and its Run again setting keeps it from running again.
Couldn’t be checked	When to use couldn’t be checked for this message, so the action didn’t run. Try again.

MCP Server tools appear when they run, as Server name: tool name. They have no Didn’t match steps, because the agent decides by itself when to use them.

Fix what you find

What you see	What to do
Almost matched	The condition is almost right. Make When to use describe what the visitor wants. See Trigger Conditions.
Didn’t match on a message that should have matched	The condition is too narrow, or it combines conditions. Remove details the action needs from When to use and make them required input fields instead. Join alternatives with “or”.
Waiting for details that never completes	Check that the input field descriptions say clearly what to ask for.
Failed	Read Received for the error. For Custom API, use the Test card to reproduce it.
Succeeded, but the answer is wrong	Check Received. If it’s missing what the agent needed, adjust the action’s response paths.