More granular observability for Vercel Sandbox
By Steven Van ·
Vercel Sandbox now reports CPU, memory, and data transfer per sandbox, queryable via CLI on Pro and Enterprise plans.
The Vercel Sandbox observability dashboard now shows detailed resource metrics, not just whether a sandbox ran. From the Observability tab, teams can monitor:
- Active CPU and CPU usage, measured in core-hours (time spent waiting on I/O, like network requests or model calls, isn't billed)
- Provisioned memory, measured in GB-hours
- Data transfer in and out of sandboxes, including package downloads and API calls
- How many sandboxes and sessions are running, and when they start and stop
Each metric can be grouped by Sandbox Name and Sandbox Session ID to trace usage back to a specific sandbox. Metrics can also be queried and visualised through the Vercel CLI, at both the team and project level, and line up with how Sandbox usage is billed. This is aimed at tracking agent workloads that spin up sandboxes at scale, right-sizing sandbox configurations, and catching unexpected data transfer. Observability is included on all plans, with manual CLI queries available on Pro and Enterprise plans. See the announcement for details.