Quality gates
Promote when green
Check workers run Lighthouse, axe, and broken-link audits on deploys. Agents read the results, fix, and redeploy — before a client ever sees the URL.
What runs
Lighthouse
Performance signals agents can act on.
axe
Accessibility findings before you pitch.
Links
Broken-link checks across the static tree.
In the agent loop
Deploy → poll GET https://api.novence.ai/v1/projects/{id}/deployments/latest until deployment.status is live or failed → read checks → fix → redeploy. Prefer MCP when available. See /for-agents.
Checks FAQ
What quality checks run?
Deployments run Lighthouse performance audits, axe accessibility checks, and broken-link checks. Agents use the results to iterate before calling the site done.
How do I read results?
After deploy, poll GET …/deployments/latest and read checksResults, or GET …/checks. MCP tools expose the same data. Do not poll project.status — it stays active.
Are checks metered?
Yes — check minutes. Free includes 30/mo. Pro 4,000. Scale 40,000. Pro/Scale soft overage $0.04 per check minute.
Why gate before pitching clients?
Failed or weak checks stay off your reputation. The agency playbook on /examples waits for green before emailing *.novence.ai preview URLs.
Ship cleaner sites
Pair checks with the agency playbook or MCP quickstart.