Team execution traces

View each team execution trace: which agents were involved, their token usage, duration, and status. Expand individual steps to see inputs, outputs, and cost breakdown.

Teams automatically hand off and delegate between agents, and the process is seamless to users. Execution traces let you go back and review "which agents were involved and how many resources each consumed" for every conversation, helping you confirm the team is dividing work as expected and making it easy to troubleshoot and optimize.

1. View execution traces

In the team page, switch to the Execution Traces view to see a list of every execution for this team. Each record includes:

Field	Description
Trace ID	The unique identifier for this execution.
Status	"Running," "Completed," "Failed," or "Timed out."
Iterations	The total number of back-and-forth exchanges between agents during this execution.
Token count	The total token usage for this execution.
Duration	The total time from receiving the message to completing the response.

2. Expand step details

Expanding a single record reveals the Step Details — which agent nodes this execution passed through in order, along with the details of each step:

Item	Description
Step number	Which step this is in the overall flow and which agent executed it.
Input	A summary of the content this agent received at this step.
Output	A summary of the content this agent produced at this step.
Cost breakdown	The token usage for this step (input tokens, output tokens, embedding tokens), letting you see exactly which agent the cost is attributed to.

Not seeing any traces? Execution traces are only generated after the team is actually triggered by a conversation. If the list shows "No execution traces yet," chat with the team through the channel first, then come back to check.

3. How to use these traces to optimize your team

• Routing not as expected: If questions are being sent to the wrong agent, go back to the canvas orchestration and adjust the connection's "trigger condition description" to be more explicit about "when to use this connection."

• Iterations or duration too high: Agents may be handing off back and forth too many times. Check for unnecessary connections, or adjust the team's "max iterations" and per-node settings.

• Cost concentrated on one agent: Use the "cost breakdown" to find the agent consuming the most tokens, and consider switching it to a lighter model or narrowing its knowledge base and tool scope.