How it works
Score → Fix → Gate → PR
The whole loop, automated. Watch a scan travel the pipeline:
◈
Scan
Real agent behavior against your live pages
score 61/100
⌖
Diagnose
Every failure mapped to a versioned class
form-field-unlabeled
⎇
Fix PR
Real code changes, opened for your review
+42 −7 · 2 files
✓
CI Gate
Merges that regress readiness get blocked
check passed · 84/100
1
ScoreA static + agent scan returns an Agent Readiness Score and a per-flow task-completion rate.
2
DiagnoseA funnel shows the exact step agents fail at, mapped to a versioned failure class.
3
FixEach failure maps to a fix: schema.org, llms.txt, ARIA/form attributes, deterministic checkout states.
4
GateA GitHub PR ships the fix; a CI check blocks future merges that regress agent-readiness.
Built to run anywhere agents do
We test against real agent behavior, so the findings hold no matter which standard (llms.txt, NLWeb, UCP) wins. Runs are async and metered — nothing sits in your request path, and our agents identify themselves honestly, never disguised as humans.