ATS Resume Optimizer: free ATS resume checker, CV builder and LinkedIn optimizer

Português: README.pt-BR.md

Check your resume the way an Applicant Tracking System (ATS) reads it, tailor it to a job description, and export a clean ATS-friendly PDF or DOCX. Works three ways:

1. Terminal app (TUI): pick your CV, paste a job, see the results live.
2. Command line (CLI): one command per task, scriptable.
3. AI agent skill for Claude Code: your AI writes the words with you; the tools check layout, keywords and truthfulness.

Optimized for ATS parsing (Workday, Greenhouse, Lever, Taleo, iCIMS, Ashby), the 7-second recruiter skim, and LinkedIn recruiter search. Zero fabrication: it never invents employers, titles, dates or numbers.

Get started in 1 minute

You need uv (one installer, handles Python for you). Then:

uv tool install "ats-resume[tui] @ git+https://github.com/paulo-amaral/ats-resume-optimizer"
ats-resume tui

That's it. Pick your resume (PDF, DOCX, TXT or JSON) in the left panel, paste the job description, and read the three tabs: ATS file, Style, Keywords.

Prefer pipx? pipx install "ats-resume[tui] @ git+https://github.com/paulo-amaral/ats-resume-optimizer".

What you can do

Real output of check on the bundled example (ats-resume new, rendered with -t modern) against this job.txt:

Senior Data Engineer
Requirements:
- Python, SQL, Spark, Kafka
- Kubernetes and Terraform
Nice to have:
- dbt, Airflow

ATS file check: en-modern.pdf
  ✓ selectable text (1906 chars)
  ✓ 1 page(s)
  ✓ email found
  ✓ standard section headings detected

Style (de-slop): 0
  ✓ no issues found

Keyword coverage vs job (ESTIMATE, not a real ATS score)
  94.0% (strong)  required_coverage=88.9%, preferred_coverage=100.0%, quantification_rate=92.3%, ...
  Missing required keywords: kubernetes
  Add missing keywords only if you truly have the skill: in Skills and in a real bullet.

The score is a labeled estimate of keyword fit. No tool can reproduce a real ATS's internal ranking, and this one does not pretend to.

Templates

Three single-column, text-based templates. All pass the built-in ATS check.

classic	modern

compact uses a serif font and tighter spacing to fit more on one page.

LaTeX. ats-resume render resume.json -o cv.tex writes LaTeX source you can edit or upload to Overleaf (set Menu → Compiler → XeLaTeX); -o cv.pdf --latex compiles it locally with Tectonic, XeLaTeX or LuaLaTeX. The LaTeX output is tuned for ATS text extraction: kerning is turned off, because kerned pairs like "AW" or "Va" extract as split words ("A WS", "V arejo"). pdfLaTeX is not used for that reason: it cannot switch kerning off, and CI caught it splitting "Texas". A test compiles the examples and fails if any word from the source is missing from the extracted text. Portuguese resumes ("lang": "pt") get localized headings and A4 paper automatically.

Use it with your AI agent

The agent writes the words with you; ats-resume checks layout, keywords and truthfulness.

Then ask, in English or Portuguese:

• "Here is my old CV (cv.pdf). Import it, check it for ATS and improve it."
• "Tailor my resume to this job: ..."
• "Optimize my LinkedIn headline and About."
• "Write a cover letter for this role."

Importing a previous CV. ats-resume import old-cv.pdf extracts what is safe to extract automatically (contact data, headline, summary, skills, language) and keeps every other section as raw text for the agent to structure, asking you when something is ambiguous. Your original stays the source of truth: ats-resume validate proves no employer, title, date or contact detail was invented. Step-by-step flow: AGENTS.md.

The skill runs a gated workflow:

Intake → Master Profile → Metric Excavation → Content → ATS Formatting
       → Keyword Tailoring → Render → Evaluate → (LinkedIn)

The Master Profile is your private record of everything you have done. Every resume is a selection from it, never an invention on top of it.

Why this one

• Honest scoring. A labeled keyword-coverage estimate, never a fake "ATS score 98/100".
• Reads your real file. Detects scanned PDFs with no selectable text, tables, images and header/footer text in DOCX, missing email, missing standard headings.
• Sourced guidance. Harvard/MIT career services, Google's XYZ formula, recruiter eye-tracking research, documented ATS parsing behavior. Myths are debunked in 07-myths-and-truth.md.
• Bilingual. English and Brazilian Portuguese: headings, A4, PT-BR keyword matching (accents included), PT-BR cliché detection, and a Brazilian market guide (LGPD, Lei 9.029/1995, Gupy).
• Private. Runs locally. Nothing is uploaded anywhere.

Myths it refuses to repeat

• "ATS auto-reject 75% of resumes." A myth from a defunct 2012 vendor. What filters people out is recruiter overload and knockout questions.
• "Hidden white-text keywords beat the bot." Modern screening flags it.
• "ATS can't read PDFs." Text-based PDFs parse fine. Two-column and image layouts are the problem.

Project layout

src/ats_resume/        CLI, TUI, checks, renderer, importer (Python package)
skills/…/SKILL.md      the AI agent workflow (start here for the skill)
skills/…/references/   ATS, bullets, keywords, LinkedIn, scoring, myths, cover letters, Brazil
skills/…/assets/       schema, templates, verb banks and cliché lists (EN + PT-BR)
tests/                 pytest suite, including render → re-parse round trips and the TUI

Develop

uv sync --all-extras
uv run pytest
uv run ruff check .

Credits

Based on artificialguybr/resume-ats-linkedin-optimizer by João Vitor Amaral (MIT). This fork adds the installable CLI, the terminal app, PDF/DOCX import and audit, three templates, Portuguese (PT-BR) support, tests and CI.

License

MIT. See LICENSE.