The engine behind the fix queue
An agent that learns from every approval.
Overlay "AI" is a script that guesses in your visitor's browser. The CloviAble Agent works the other way: it studies real findings across real sites, proposes a fix, and — this is the part that matters — learns from what humans approve and reject. Every decision in the fix queue trains it. The more sites it fixes, the more complex the sites it can fix.
The loop is already running — live numbers
8,969real findings harvested into the learning storeharvest_findings, live DB
595scan runs feeding the flywheelharvest_runs, live DB
6fix patterns confirmed across multiple independent siteslearned_patterns.json
emergingthe agent's own published confidence tier — it grades itself honestlylearned_patterns.json meta
The honest part: we publish the agent's confidence tier and it starts at the bottom. A pattern only graduates when it's held up on multiple independent sites, and human-judgment items are never auto-passed. "AI-powered" is easy to say; a learning system that shows you its own report card is the version we'd buy.
Where the agent goes next
- Approve/reject as first-class training signal — every fix-queue decision updates pattern confidence per rule, per platform (WordPress vs Shopify vs custom).
- Complexity tiers — from single-page fixes to cross-template and SPA repairs, unlocked as pattern confidence graduates.
- Your site's own memory — the agent remembers your stack's quirks, so fix #40 on your site is better than fix #1.