Documentation/Simulation Loop

Simulation Loop

Understanding the simulation lifecycle helps you optimize test runs and interpret results.

Phase 1: Initialization (0-30s)

  • Agents are spawned with assigned personas
  • Initial page loads and screenshots captured
  • Baseline metrics established

Phase 2: Warm-up (30s-2m)

  • Agents explore the environment freely
  • Navigation patterns are learned
  • Common paths are identified

Phase 3: Task Execution (2m-10m)

  • Agents attempt to achieve their assigned goals
  • Interactions are logged in detail
  • Friction points are identified
  • Success/failure rates tracked

Phase 4: Teardown (10m+)

  • Agents complete their sessions
  • Data is aggregated and analyzed
  • Reports are generated
  • Heatmaps and funnels created

Configuring Simulation Duration

crowdi run --url https://your-product.com --duration 15m

Longer simulations provide more data but cost more credits.

Real-time Monitoring

Watch agents in action with Live Pulse:

crowdi run --url https://your-product.com --live

Opens a real-time dashboard showing agent actions as they happen.

Last updated: Today

Edit this page on GitHub →
Crowdi - Synthetic User Testing