Vercel always-on tracing for production and preview traffic Targets Real-World Debugging.

Vercel is giving developers a clearer view of what happens after an application reaches real users. Its new always-on tracing capability samples production and preview traffic, allowing teams to investigate live requests without first reproducing a problem in their own browser.

The feature, detailed in an official changelog entry, continuously collects distributed traces according to rules configured by the user. Those traces can help developers follow request flows, identify slow operations and narrow down where failures occur across an application’s runtime components.

That distinction matters for intermittent bugs. Problems tied to a particular route, data payload or production environment can be difficult to recreate locally. By retaining sampled traces from actual traffic, Vercel gives teams evidence from the request that failed rather than asking them to reconstruct its conditions afterward.

Despite the “always-on” label, Vercel does not begin collecting everything automatically. A developer must first create at least one sampling rule. Each rule specifies a trace rate and can apply to all environments, production only or preview deployments only.

Rules can also be limited to a path prefix such as /checkout. That gives teams a practical way to focus tracing on important or problematic areas instead of sampling an application uniformly. A business investigating abandoned purchases, for example, could increase coverage for its checkout routes while leaving less critical traffic at a lower rate or untraced.

The available controls include:

• Environment scope: All, Production or Preview.
• Trace rate: The proportion of matching live requests to collect.
• Optional path scope: A prefix used to target routes such as /checkout.
• Opt-in collection: No traces are collected until a sampling rule is added.

This setup is also Vercel’s mechanism for controlling cost. The company says customers pay only for the traffic they choose to trace, although the announcement does not provide plan requirements or specific pricing. Teams should therefore review their selected sampling rates and route coverage before enabling broad production collection.

Vercel already offers session tracing, but that system captures requests generated from a developer’s own browser session. It is useful when someone can deliberately walk through a workflow and trigger the behavior under investigation.

Always-on tracing addresses the cases that approach misses. It samples traffic from real users in production and from people testing preview deployments. A developer no longer needs to know the exact reproduction steps before useful diagnostic information can be collected.

The production and preview split should be particularly useful for small teams operating without a dedicated reliability function. Preview tracing can expose regressions while a change is still under review, while production tracing can reveal issues caused by real traffic patterns. Together, the two environments provide a more continuous path from pre-release testing to live operation.

There are still trade-offs. Sampling means not every request will necessarily produce a retained trace, and narrow path rules may omit activity outside their configured scope. Teams will need to balance diagnostic coverage against collection volume, then adjust the rules when an incident points to a specific route or environment.

The release fits Vercel’s broader effort to bring production operations closer to the workflows developers already use. The company has also been extending production-aware actions into collaboration software through Vercel Agent in Slack. Always-on tracing focuses on the underlying request evidence, giving teams another source of context when a deployment behaves differently under real-world conditions.

What is Vercel always-on tracing for production and preview traffic?

It is a distributed tracing feature that continuously samples real requests from Vercel production and preview deployments. Developers can use the collected traces to investigate failures and performance issues without reproducing each request in their own browser.

Does Vercel collect traces automatically?

No. Collection begins only after a user creates a sampling rule. The rule defines a trace rate, an environment and, optionally, a path prefix to target.

How is always-on tracing different from session tracing?

Session tracing captures requests originating from the developer’s own browser session. Always-on tracing instead samples live production or preview traffic, making it better suited to intermittent issues experienced by real users.