Vercel Observability tracks eve agent runs

Adding agent-specific observability directly into the deployment platform solves a major developer headache, as tracing LLM calls and tool executions usually requires complex OpenTelemetry wiring.

• –Zero-Config dashboard: The Agent Runs tab auto-populates for any eve project without requiring manual OpenTelemetry instrumentation or instrumentation.ts setup.
• –Developer vs. Business modes: Splitting telemetry into raw JSON views for debugging and plain-English summaries for audit logs caters perfectly to both engineering and compliance needs.
• –Retention limits: Short retention windows (12 hours on Hobby, 1 day on Pro, 3 days on Enterprise) feel restrictive, likely pushing serious teams to buy the Observability Plus add-on or export spans to custom destinations like Braintrust or Datadog.
• –Error correlation: Capturing and associating runtime crashes with the exact agent turn or tool execution helps resolve the notoriously difficult problem of tracing errors back to specific model states.