More granular observability for Vercel Sandbox
Vercel just gave AI agents a cost tracking superpower. Now you can see exactly which sandbox is burning through your compute budget.
Why it matters
As AI agents spin up sandboxes at scale, granular observability becomes a competitive necessity. Vercel's new metrics let builders attribute costs to specific workloads and right-size infrastructure—critical for managing runaway agent spending.
The key facts
11 to knowNew observability dashboard tracks Active CPU (core-hours), Provisioned Memory (GB-hours), Data Transfer, and Running Sandboxes/Sessions
Metrics groupable by Sandbox Name and Sandbox Session ID for drill-down analysis from aggregate to individual sandbox level
Available on all plans; manual queries available on Pro and Enterprise
Designed specifically for tracking agent workloads that create sandboxes at scale
Cost attribution now tied directly to sandbox usage billing model
New granular metrics: Active CPU (core-hours), Provisioned Memory (GB-hours), Data Transfer, Running Sandboxes/Sessions
Metrics grouped by Sandbox Name and Sandbox Session ID for drill-down attribution
Available on all plans; manual queries on Pro and Enterprise
Feature explicitly targets agent workloads that create sandboxes at scale
CLI query and visualization support included
Cost attribution aligned with billing model
Go to the source
Vercel Blogvercel.com
Publisher excerpt: observability now includes detailed resource metrics, giving you deeper visibility into how your sandboxes consume compute and networking.Vercel Sandbox From the in your dashboard, you can now monitor:Observability tab Each metric can be grouped by Sandbox Name and Sandbox Session ID, so you can…