Skip to main content

Analytics

Foil provides comprehensive analytics to understand your AI application’s performance, costs, and quality.

Dashboard Overview

The main dashboard shows key metrics at a glance:
MetricDescription
Total RequestsNumber of traces in the time period
Success RatePercentage of successful completions
Avg LatencyMean response time
Active AgentsAgents with recent activity
Alert CountActive alerts requiring attention

Available Analytics

The dashboard includes charts and breakdowns for:
  • Requests over time — volume trends with agent breakdown
  • Success/failure rates — error trends and patterns
  • Latency distribution — response time buckets and p50/p95/p99 percentiles
  • Token usage — input/output token consumption over time
  • Cost breakdown — costs by model and time period
  • Error analytics — errors by type and error rate trends
  • Tool usage — which tools are used by which agents
All analytics support filtering by date range, agent, and time granularity (hourly, daily, weekly, monthly). These metrics are also available via the REST API.

Best Practices

Average latency can hide outliers. p95 shows what your slowest users experience.
Monitor costs closely, especially when testing new prompts or models.
Regular comparisons help identify regressions quickly.

Next Steps