That monolithic frontend you've been avoiding? You can open it again.
Targeted exports from Frontinus to your LLM
Every finding pinpointed to the exact line and column
Scan & Map - Triage & Classify - Fix & Verify
Track your project health session after session. Launch targeted LLM missions — from dead code cleanup to specificity refactoring — each one a structured context pack, one click away. See your score rise, review what you've fixed, and always know what's next.
Coming soon: LLM missions, GitHub workflows, full audit timeline.
First 100 subscribers — locked at this price forever.
No credit card required.
You can upload a monolithic HTML file with embedded <style> blocks, or separate CSS and HTML files independently. JavaScript is analyzed as additional context when uploaded alongside HTML — it powers cross-reference detection for JS-toggled classes and DOM interactions. No build step, no configuration, no node_modules. Upload, scan, work.
Frontinus is built around modules. The CSS Module — available now — targets what most production codebases actually contain: monolithic files, embedded styles, and years of accumulated decisions. Legacy is the real problem. React and Tailwind modules are in active development — same structure, dedicated engines. Not a port.
Linters catch syntax and rule violations. Frontinus maps relationships — between your CSS, HTML, and JS — and packages findings as structured context packs for your LLM. The goal isn't to flag errors. It's to give Claude or GPT exactly what they need to fix them without guessing.
After the 14-day trial, you'll need to subscribe to keep working — no automatic charges. If you don't, your account is locked but nothing is deleted: your projects stay exactly as you left them. Subscribe anytime to pick up right where you left off.
Cancel anytime — no charge until you resubscribe. Your projects aren't touched: everything stays exactly where you left it. When you're ready, resubscribe (monthly or annual) and you're back in instantly — no re-upload, no re-setup.
Your project is stored in an isolated environment tied to your account. You can resume your work anytime and delete it whenever you choose. Only essential cookies are active by default — analytics and marketing require your explicit consent.
Only you can access your source code through the application. Our team may access project data only when necessary to provide support, fix bugs, or ensure service reliability — never for any other purpose. Your code is never used to train AI models.
Frontinus exposes its API surface as a programmatic catalog so LLM agents can read it without parsing screenshots. To obtain the full list of endpoints, schemas, and notes, use one of the three channels below.
javascript_tool):
JSON.stringify(window.frontinus.apiSurface(), null, 2)
read_page / get_page_text):
document.getElementById('frontinus-api-surface-data').textContent
<script type="application/json" id="frontinus-api-surface-data">.
Read its textContent and JSON.parse it.
{
"path": "/check-css",
"method": "POST",
"purpose": "Validate CSS with stylelint, multi-file aware",
"tier": "core",
"schema_version": "3.13.0"
}
Snapshot the current working state at the end of each refactor phase. Granular rollback to any saved point — useful when a later change breaks something and you want to recover an earlier good state without losing other phases.
Append-only log of every action on this project. Export as JSON or CSV — the SHA256 fingerprint certifies integrity. The trail is the value here: there is no approval workflow to go through.
Upload screenshots or short videos as a visual baseline of the app pre-refactor. Compare side by side with post-refactor results to catch visual regressions that static analysis cannot detect from code alone. Tag each asset as before, after, or reference.
Click a thumbnail to load it (alternates between left and right) — or use the ← / → buttons to choose explicitly.
A document a client can hand to an LLM they choose, with access to the project files, to check a quote against the real state of the project. "Does this scenario match reality?" is verifiable fact-by-fact; "is the quote correct?" is not — the report says so explicitly and leaves that judgment as a placeholder.
The Causality Engine is a local companion tool that diagnoses why your layout changes — constraint propagation, cascade conflicts, responsive breakpoints, DOM mutations — and produces a structured report ready for LLM consumption.
Don't have the plugin yet? A distributable package is coming soon. In the meantime, contact the team for early access.