Paste your CV + a job posting. Get the exact keywords you're missing, what to add, and what to cut — with proof quoted from your own CV.
English · Français · العربية (RTL) · works on mobile · free forever
Most companies run CVs through an ATS (Applicant Tracking System) before a human ever sees them. One missing keyword — TypeScript, PostgreSQL, anglais courant — and a qualified candidate is silently filtered out. People get rejected by software for formatting they never knew was wrong.
CV Lens shows you exactly what that filter sees.
| 🧾 Evidence-based keyword table | Every must-have and nice-to-have from the posting: FOUND / PARTIAL / MISSING — each with a short quote from your CV (or an honest "no evidence") |
| 📍 Keyword placement | For each keyword: does it live in your Experience section or only in your skills list? Real ATS parsers treat these very differently. Synonym-aware (JS = JavaScript, CI/CD = Continuous Integration…) |
| 🎯 Must-haves first | Required keywords are separated from nice-to-haves and sorted worst-first, with an "X/Y must-have keywords matched" summary |
| 📊 Match score + 5 sub-scores | Must-haves · keywords · experience · education · ATS formatting — with an honest verdict, not flattery |
| 🔥 Before / after tracking | Each analysis is compared with your previous one on the same device — see your score climb |
| ✍️ AI bullet rewrites | One click rewrites your 4–6 weakest bullets using the job's keywords — grounded in your CV only, unknown numbers stay as [X%] placeholders, never invented |
| ✅ Add | Concrete additions with suggested wording — always phrased "only if true for you" |
| ✏️ Improve | Before → after rewrites of your actual lines |
| ❌ Remove | What's hurting you for this job specifically |
| ⚡ Rule-based ATS checks | Contact info, section headings, action verbs, quantified results, single-column layout, dates, employment gaps, verb variety, overlong bullets, images and repeating headers/footers detected inside the PDF — deterministic facts, not AI guesses |
| 🔗 Smart job-link handling | Paste a job URL and press Analyze — it fetches the posting automatically (server-side, no CORS walls) |
| 📄 Client-side document parsing | PDF / DOCX / TXT read locally with pdf.js + mammoth.js — the file never leaves your device |
| 🕘 History | Last 10 analyses kept in localStorage — reopen any past report (stays on your device) |
| 📋 Export | Copy the full report, download a PDF report (jsPDF), print, or share a score card image for social media |
- Your CV file never leaves your browser — parsing happens locally; only plain text you submit goes to the analysis API
- Nothing is stored or logged — no accounts, no database, no tracking
- No keys in the client — the AI key lives only in a serverless function (Vercel env secret), never in shipped JavaScript
- Rate-limited API (10 analyses / hour / IP) to prevent abuse
Browser (GitHub Pages) Serverless (Vercel, free tier)
┌───────────────────────────┐ ┌─────────────────────────────┐
│ PDF/DOCX parsed locally │ POST │ rate limit → OpenRouter AI │
│ rule-based ATS checks │ ────────► │ (key = env secret, never │
│ render scores & evidence │ ◄──────── │ shipped to clients) │
└───────────────────────────┘ JSON │ JSON validated + auto-retry │
└─────────────────────────────┘
Hybrid accuracy: deterministic rules in the browser + a strict evidence-based AI prompt (temperature 0.2, JSON schema, model fallback chain) that is explicitly forbidden from inventing experience.
Zero-framework vanilla JS · Tailwind (CDN) · pdf.js · mammoth.js · jsPDF · Vercel serverless functions · OpenRouter AI · EN/FR/AR i18n (RTL)
git clone https://github.com/hajar-benhadj/ats-cv-checker.git
cd ats-cv-checker
python -m http.server 8080 # open http://localhost:8080The public API endpoint is already live; to self-host the analysis too, deploy the api/ functions to your own Vercel and set OPENROUTER_KEY.
MIT © Hajar Benhadj — use it, learn from it, build on it.