TestingAI Guide

TestingAI runs real browser navigation using Playwright/Chromium on the backend. It spawns multiple agents, records videos, streams execution logs, and produces a visual/UX analysis summary.

Where to run TestingAI
  • Open your Integration dashboard
  • Go to the TestingAI tab
  • Set the Target URL (the site TestingAI should navigate)
  • Click Run (the UI will show Running… with an elapsed timer)
Controls (Agents, Pages, Clicks)

TestingAI uses guardrails to keep runs predictable. Common controls:

  • Agents: number of parallel browser agents (capped for stability)
  • Max Pages: how many distinct pages an agent can visit
  • Max Clicks: how many interactions an agent can perform
  • Accessibility: optional accessibility scan (when enabled)

Under the hood, the API call looks like:

POST /ai/swarm/run
Authorization: Bearer YOUR_TOKEN
Content-Type: application/json

{
  "integrationId": "INTEGRATION_ID",
  "agents": 8,
  "targetUrl": "https://www.example.com",
  "maxPages": 3,
  "maxClicks": 10,
  "includeAccessibility": true
}
Outputs (Videos, Logs, Visual Analysis)
  • Execution Logs: streamed in real-time (SSE) and shown in the dashboard
  • Agent Videos: one video per agent (download/preview)
  • Visited Pages: list of pages accessed during the run
  • Visual Analysis: average UI score and per-agent scores
Important: TestingAI does NOT count as real users

TestingAI traffic is tagged with a distinct User-Agent so it is excluded from real-user telemetry:

User-Agent: TestingAI Sublyzer/agent-1

This prevents TestingAI runs from affecting charts like IPs by Country and Operating Systems.

Troubleshooting
  • Run fails immediately: verify the target URL is valid and publicly reachable
  • No navigation happens: the environment may lack a working headless browser; TestingAI can fallback to static analysis
  • No logs: verify the SSE stream is reachable and not blocked by proxies

See also: Troubleshooting

Learn about SwarmAI →