About this block

Agent Run Trace is the step-by-step timeline the assistant shows while it works a request. A vertical rail lists five steps: interpreted the request, queried renewals, searched support tickets, cross referenced both result sets, and a final drafting step still in progress. Each step pairs a plain language label with a tool badge (Plan, SQL, Search, Analyze, Compose), a right-aligned duration, and a detail line; the SQL and Search steps render their exact query in a scrollable monospace strip. The footer is honest about depth: every run is logged step by step, and the raw log holds each query and the exact rows the assistant read.

Steps is one typed array; the status field flips the rail marker between a filled check and a spinning loader, and the mono flag swaps the plain detail line for the code strip. Showing the literal query beside a readable summary is what lets a reviewer confirm the agent did what it claimed.

When to use

Reach for this block inline in the conversation while an agent answers a multi-step question, or in a run detail drawer opened from a job log. Wire the steps to your agent event stream so the last row stays in the running state until the tool call returns.

A natural flow around it on an Application Pro page:

Before

• Prompt composer

After

• Citations panel
• Activity feed

What you can display

One strong use is the live agent reasoning trace. Other step-timeline shapes:

• Automation run log. Workflow steps with a status badge and the record each step touched.
• Deploy pipeline. Build, test, and release stages with per-stage duration and output.
• Import job trace. File parse, validate, and write steps with a row count on each detail line.

Tip: keep the running step visible at the bottom so users read the agent as working, not stalled.