Traces

Debug individual executions. See exactly what happened.

Viewing Traces

Each trace shows:

Workflow ID
Duration and cost
Success/failure status
All spans (steps) in order

Trace Detail View

Click any trace to see the full breakdown:

research_pipeline - 4.2s, $0.67, SUCCESS

├─ search_agent (1.2s, $0.12)
│  └─ openai.embeddings
├─ analyze_agent (2.1s, $0.35)
│  └─ anthropic.claude-3-opus
└─ write_agent (0.9s, $0.20)
   └─ openai.gpt-4o

Filtering

Filter by any dimension:

Filter	Example
Workflow	`workflow_id = research_pipeline`
Customer	`tenant_id = acme_corp`
Status	`status = error`
Cost	`cost > 0.50`

Debugging Failures

Failed traces show exactly where things broke:

research_pipeline - FAILED

├─ search_agent (1.2s, $0.12) ✓
├─ analyze_agent - RateLimitError ✗
└─ write_agent (skipped)

Next Steps

Monitor — Set up alerts
API Reference — Programmatic access