jobPilot

A local-first job search tool that finds, ranks, and tailors job matches using your resume and Claude AI. Runs entirely on your machine — your resume data never leaves your computer.

What it does

• Pulls job listings from Adzuna (and optionally other boards) based on your search criteria
• Scores and ranks matches against your resume using AI, surfacing the jobs most worth your time
• Tailors your resume for a specific role with one click — highlights relevant experience, flags gaps, suggests edits
• Tracks applications with a lightweight status tracker (New → Applied → Interviewing → Offer → Rejected)
• Edits your resume in-place — update contact info, summary, skills, and experience without leaving the app, with an AI-assisted summary rewrite helper

How it works

jobPilot runs a local FastAPI server and opens in your browser. A system tray icon keeps it running in the background and auto-refreshes job listings every 12 hours.

All data — your resume, job listings, match scores — lives in ~/.jobpilot/ on your machine. LLM calls go directly to the Anthropic API using your own key.

Setup

Prerequisites

• Python 3.11+
• An Anthropic API key
• An Adzuna API key (free tier works)

Install

git clone https://github.com/drewmerc302/jobPilot.git
cd jobPilot
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install -e .

Configure

Create ~/.jobpilot/.env:

ANTHROPIC_API_KEY=sk-ant-...
ADZUNA_APP_ID=your_app_id
ADZUNA_APP_KEY=your_app_key

Run

python -m jobpilot

The app opens in your browser at http://127.0.0.1:8765. On first launch, a setup wizard walks you through uploading your resume and configuring your search.

First run

1. Upload your resume — PDF or DOCX. Claude extracts your experience, skills, and contact info.
2. Confirm the extraction — review and correct anything that looks off.
3. Set up your search — location, radius, keywords, target companies.
4. Get your first matches — ranked by fit, ready to review.

Features

Job matching

Jobs are scored against your resume on relevance, seniority fit, and keyword overlap. The matches list shows score, company, title, salary, and location at a glance.

Job detail analysis

Every job detail page shows an AI breakdown of the role against your profile: why it fits, the key requirements extracted from the description, and interview talking points specific to your background.

Resume tailoring

Click Generate Tailored Resume on any job. Claude produces a diff of your resume bullets — current vs. suggested rewrites with a rationale for each change. Check the edits you want to apply, then click Generate PDF to produce a tailored resume ready to send. Your stored profile is unchanged.

Interview Q&A Guide

One click generates a set of likely interview questions for the role with suggested answers drawn from your specific experience. Available on every job detail page (~$0.03).

Resume editor

The Resume page lets you edit your stored profile directly — the data that drives all matching and tailoring:

• Add new experience in plain language at the top; Claude structures it automatically on save
• Edit contact info, summary, skills, and structured work history
• AI summary rewrite: describe your recent accomplishments, get a polished 2–5 sentence draft

Application tracking

Track each role through your pipeline via the status dropdown on the job detail page (New → Applied → Interviewing → Offer → Rejected).

Contextual tooltips

Hover over any button, column header, or form label to get a plain-language explanation of what it does and what it costs.

Documentation

A full annotated user guide is available at docs/user_guide/. Build the PDF locally:

cd docs/user_guide
typst compile user_guide.typ user_guide.pdf

Requires Typst. The source references screenshots from screenshots/ (gitignored — see SHOT_LIST.md for what to capture).

Packaging

jobPilot uses Briefcase for native packaging.

Vendored Typst binaries

Resume PDF generation uses Typst. Binaries are vendored at build time (~40 MB each, gitignored) into src/jobpilot/resources/typst/<arch>/. Run once before building or developing:

# All four targets (arm64+x86_64 macOS, linux, windows):
python scripts/fetch_typst.py

# Just the host arch (faster for dev):
python scripts/fetch_typst.py --current

The bash (scripts/fetch_typst.sh) and PowerShell (scripts/fetch_typst.ps1) wrappers do the same thing.

macOS:

briefcase build macOS
briefcase run macOS

universal_build = false in pyproject.toml — produce one bundle per host arch and ship both, or set true if you have an arm64+x86_64 Python.

Windows — requires a Windows machine or CI. See docs/WINDOWS_BUILD.md for full instructions including a ready-to-use GitHub Actions workflow.

Cost

All AI features use your Anthropic API key and are billed to your account. Rough estimates:

• Resume extraction (one-time): ~$0.05
• Match scoring per run: ~$0.01–0.03 depending on result count
• Tailor analysis + resume PDF per job: ~$0.03
• Interview Q&A Guide per job: ~$0.03
• AI summary rewrite: ~$0.03
• New experience parse: ~$0.01

A cost meter in the nav bar shows your running total.

Tech stack

• Backend: FastAPI + uvicorn
• Frontend: Jinja2 templates + HTMX (no build step)
• Database: SQLite
• AI: Anthropic Claude (Sonnet for tailoring/summaries, Haiku for extraction)
• Job data: Adzuna API
• Packaging: Briefcase (macOS app bundle, Windows MSI)
• System tray: pystray + Pillow

Data & privacy

Everything stays local. Your resume is stored in ~/.jobpilot/profile.json. Job listings are cached in ~/.jobpilot/jobpilot.db. The only outbound calls are to the Anthropic API (for AI features) and Adzuna (for job listings).

License

MIT