Skip to main content
Recaps are AI-generated performance reports that dig through your agent data and surface what you’d otherwise miss - emerging failure patterns, cost anomalies, shifting user behavior, and the specific runs behind each trend. Configure them on a daily, weekly, or monthly cadence.

​
How it works

  1. Data collection - queries your agent’s runs, errors, costs, and user interactions from the configured time period
  2. AI analysis - an AI agent explores the data end-to-end, identifying trends, anomalies, and noteworthy changes
  3. Evidence linking - each finding is backed by links to the specific runs, sessions, and patterns that support it
  4. Delivery - recaps are sent to your Slack channels and appear in your dashboard

​
What’s inside a recap

Every recap includes:
  • Metrics comparison - key numbers (run volume, failure rate, cost, feedback) compared against the previous period so you can spot shifts instantly
  • Trends of interest - the AI highlights what’s changing and why it matters, whether it’s a spike in errors, a drop in user satisfaction, or a model that’s costing more than it should
  • Issues by severity - problems categorized as high, medium, or low, each with supporting user quotes and linked runs
  • Topic breakdown - what users are asking about most, with percentage breakdowns and example prompts
  • Highlights - the wins: improved success rates, efficiency gains, tools performing well

​
Configuring your cadence

Set your recap frequency from the Integrations page or through Slack:

​
Dashboard highlights

Recaps also appear as highlights at the top of your overview dashboard. Click any highlight to drill into:
  • A written analysis of what changed and why
  • The key metrics behind the finding
  • Links to the specific failures, runs, sessions, or patterns involved
  • Example traces illustrating the trend

​
Slack delivery

With the Slack integration configured, recaps are delivered automatically to your subscribed channels on your chosen schedule.