Vercel Agent adds Critical-only filtering before webhook paging

Vercel added alert validation on October 26, allowing Vercel Agent to investigate monitor alerts before they reach a paging system. When Critical Only is enabled for a webhook destination, only alerts the agent classifies as Critical are sent to that destination. The classification uses logs and metrics, and teams can add monitor‑specific context to guide the investigation.

The resulting delivery decision depends on the agent’s assessment. The announcement does not supply an independent accuracy evaluation or establish that every important incident will be classified correctly. OpenTools has not tested the feature against production incidents.

The release’s dashboard workflow starts in a monitor’s notification settings: create or link a webhook destination for the paging system, then select Critical Only next to that destination. The same change is available through vercel monitors notifications; add --critical‑only when linking or updating a webhook destination.

Vercel’s webhook documentation says an endpoint receives alerts.triggered events for its linked monitors, including their trigger conditions and project scope. The same endpoint can be linked to multiple monitors. A webhook with no linked monitors receives no alert events, so the endpoint’s relationship to the monitor is part of the routing configuration.

The [documentation](https://vercel.com/docs/alerts/configure‑alerts#configure‑webhooks) explicitly says Agent’s Critical assessment can delay a notification relative to the monitor’s overall severity threshold. Teams deciding where to use the setting therefore need to consider both the classification gate and the extra time before delivery; the notice does not give a fixed investigation‑latency guarantee. There is also a migration detail for existing destinations. Account‑wide alert webhooks continue receiving all alert events until their configuration changes. When an account‑wide webhook is linked to a monitor, Vercel limits future delivery to the monitors linked to that endpoint. That is a change in the set of events routed to the endpoint, separate from choosing whether it should receive only Critical alerts.

Alert validation is available on Pro and Enterprise plans with Observability Plus, according to Vercel’s release. Observability Plus includes 10 Agent investigations per billing cycle, with additional investigations billed separately. The included count is per billing cycle, rather than a per‑monitor or daily allowance stated in the announcement.

The release separately prices custom monitors at $0.0035 per GB of data scanned and says built‑in alerts remain included with Observability Plus. That scanned‑data rate and the investigation allowance describe different billing items. A workload estimate needs the relevant monitor data volume and investigation usage, rather than treating either published number as a flat price for all alert handling.

AI‑generated conceptual illustration by OpenTools Team: an inspection lens, alert cards and a phone represent review before notification. This is illustrative artwork, not Vercel’s interface, an incident test or evidence of classification accuracy.