Swarm QA
Swarm QA is xyva.ai's multi-agent scanning system. It deploys up to eight specialized AI agents in parallel, each targeting a different quality dimension — from broken links to accessibility compliance to security headers. Results are aggregated into a single score and findings dashboard.
How It Works
- You select a target URL and a preset (or configure agents manually in Expert Mode).
- Swarm launches the selected agents simultaneously, each crawling the target independently.
- Agents report findings in real-time to the Live dashboard.
- When all agents finish, xyva calculates a composite quality score (0-100) and stores the run in history.
INFO
Swarm QA itself is part of the paid core product. Advanced scheduling, webhooks, and private deployment operations are available on request.
Dashboard Tabs
The Swarm QA page is organized into four tabs:
| Tab | Purpose |
|---|---|
| Live | Real-time view during an active scan. Shows agent progress, finding count per agent, and elapsed time. |
| History | Chronological list of all completed runs. Each entry shows the score, trend indicator, finding counts by severity, and an expandable detail view. |
| Findings | Aggregated findings across all runs. Recurring issues are marked with a repeat indicator. Each finding has an "Analyze with AI" button for deeper investigation. |
| Schedule | Configure recurring scans at intervals from 30 minutes to 24 hours. Select a preset, enable auto-refresh, and set up webhook notifications. |
Agents at a Glance
| Agent | Focus area |
|---|---|
| Link Patrol | Broken links, redirect chains |
| HTTP Guard | Security headers, cookies, mixed content |
| A11y Scout | WCAG/BFSG accessibility compliance |
| Smoke Flow | Page loads, console errors, rendering |
| Perf Sentinel | Core Web Vitals, load performance |
| SEO Recon | Meta tags, structured data, Open Graph |
| Form Fuzzer | Input validation, XSS, injection |
| API Health | Endpoint availability, response times |
See Agent Types for full details on each agent.
Sub-Pages
| Page | Content |
|---|---|
| Agent Types | All 8 agents with detailed check lists |
| Presets & Expert Mode | Quick Smoke, Standard, Deep Audit configuration |
| Findings & Scoring | Severity, action categories, AI enrichment, score formula |
| History & Comparison | Run listing, trends, side-by-side diff |
| Visual Regression | Pixelmatch comparison, threshold config |
| Schedule & Webhooks | Recurring scans, Slack/Teams/Discord notifications |
| PDF Export | Single-run and aggregate reports |
| Authentication | Login flow for protected pages |
| CLI Runner | CI/CD integration, exit codes, JSON output |