JobHound

It fetches. You interview.
Companies use AI to filter candidates. JobHound gives you AI that hunts companies.

---

JobHound is an AI-powered job search command center that runs inside any AI coding CLI (Claude Code, OpenCode, Gemini CLI, Codex, Qwen). Chat with it once about who you are and what you want — it calibrates its search agents to your experience, industries, titles, and job boards, then hunts on a schedule.

What the hound does

🐾 Learns you first	Conversational onboarding: paste your resume or just talk about your experience — it builds your CV, profile, and search config from the chat
🎯 Hunts on your terms	Tell it in plain language: "only Staff+ roles posted this week, fintech, remote US, nothing under $140K" — it encodes your rules and applies them on every sweep
📡 Sweeps live boards for $0	sweep-live-boards.mjs hits Greenhouse + Ashby public APIs directly — zero LLM tokens — verifies liveness, dedupes, and drops fresh leads into your pipeline
⭐ Scores every offer	Multi-archetype fit scoring, salary filters, repost detection, ATS keyword match
📄 Tailors your CV per company	Generates ATS-optimized PDFs and cover letters from your single source-of-truth CV
✍️ Fills applications with you in the loop	apply-assist.mjs opens the real ATS form in a browser, prefills everything from your profile, uploads the right CV — you review and click Submit
🗂️ Runs your Apply Hub	One status board for every prepped application, with per-role packages and run commands

Quick start

git clone https://github.com/aabrole/jobhound.git
cd jobhound
npm install

# open the folder in your AI coding CLI (e.g. `claude`) and just say hi —
# the hound walks you through setup by chatting: your resume, target roles,
# industries, boards, and exactly how you want finds presented.

On first launch there is nothing to hand-edit: onboarding builds cv.md, config/profile.yml, and portals.yml from the conversation.

Teach it how to hunt

The core idea: your filters are conversations, not config files. During onboarding (and any time after), tell the hound things like:

• "Only postings from the last 7 days — flag reposts"
• "I have 4 years of experience; skip anything asking for 8+"
• "Include Product Manager, exclude Project Manager"
• "Healthtech and climate only. No crypto."
• "Add these five company boards and this niche job board"
• "Every Monday give me a top-10 ranked table with salary and links"

It translates those into portals.yml, config/sweep.yml, and your profile — and honors them on every scan, sweep, and report.

The manual controls

Everything is also runnable directly:

node sweep-live-boards.mjs        # zero-token sweep of your calibrated boards
node scan.mjs                     # full portal scan
node apply-assist.mjs "<url>"     # human-in-the-loop application autofill
npm run pdf                       # tailored CV → PDF
npm run doctor                    # health check

See ARCHITECTURE.md for the full system map and AGENTS.md / CLAUDE.md for how the AI operates it.

Credits

JobHound is a spinoff of career-ops by Santiago Fernández de Valderrama (MIT) — a phenomenal foundation, battle-tested across 740+ evaluated offers. JobHound extends it with the zero-token live-board sweeper, the review-gated application autofill assistant, the Apply Hub command center, conversational filter calibration, and a very good dog.

Mascot: original 8-bit pixel art, generated from ASCII grids by docs/brand/gen.mjs.

License

MIT — original code © Santiago Fernández de Valderrama, JobHound additions © Aman Abrole.