Vercel's trace search lets developers and coding agents start investigating a request failure without first knowing its trace ID. The CLI update adds vercel traces search, which returns individual steps within requests, called spans. By default it retrieves up to 100 spans from the past hour, newest first. Filters can narrow that set to errors, a service or a deployment; a duration query can find slow steps. JSON output gives scripts and agents structured records to inspect, moving the initial search for a failing or slow operation into the terminal.
Key takeaways · AI-distilled
By default, vercel traces search returns up to 100 spans from the last hour, newest first, so you can start investigating without first locating a request or trace ID.
Dedicated flags filter by environment, service, span name, status, deployment, request, trace or span ID, and --since/--until narrow the window; for example, --status error --since 1h lists the last hour's error spans.
Fields without their own flag, such as duration or span and resource attributes, go through --query using Vercel's supported subset of KQL, e.g. 'span.duration > 2000' for spans slower than two seconds. The command needs the latest CLI via vercel upgrade.