OPEN SOURCE Bring your own coding agent — your data stays local

Your job hunt,
as mission control.

A tiny, dependency-free toolkit and a beautiful local dashboard for running a serious job search. Discover roles, score them for fit, track your pipeline, tailor CVs, and keep every recruiter email in one place — driven by your own coding agent.

$ git clone …/huntbench && python3 jobsdb.py serve
View on GitHub
No accounts No API keys Python 3 stdlib only MIT licensed
127.0.0.1:8765
Huntbench dashboard — status-grouped job list with fit scores
Bring your own agent

The smart parts are done by your LLM.

Huntbench is the deterministic engine and the local UI. Your coding agent — Claude Code, Codex, Cursor — does the finding, tailoring, and form-filling using tools you already have. Nothing phones home.

1

Clone & serve

One command. python3 jobsdb.py serve opens the dashboard with a live demo dataset — no install, no signup, no keys.

2

Say “set me up”

Open the folder in your agent. It reads AGENTS.md, interviews you, and writes your profile and master CV. The dashboard detects it and you're in.

3

Hunt from one board

Scan ATS boards, triage the pipeline, generate tailored CVs, and prep applications — your agent fills forms and stops before submit for your review.

The dashboard

Everything about the hunt, in one calm place.

A light, Linear-style workspace — a status-grouped list, a drag-to-move board, and an inbox that turns recruiter email into a timeline.

A pipeline that scores itself

Every role lands with a 1–5 fit score from your target roles and title filters. Rows carry the seniority, stack, location and work-mode at a glance — grouped by status so you always know what's next.

  • New → Applied → Interviewing, at a glance
  • Region & work-mode filters in the sidebar
  • A ⌘K palette to jump or run anything
127.0.0.1:8765
Status-grouped job pipeline with 1–5 fit scores

Drag your way down the funnel

A Kanban board with a column per stage — New, Shortlisted, Applied, Screening, Interviewing. Drag a card to move it; the fit badge and company travel with it. It's the same pipeline, seen the way you think about it.

  • Drag-to-move between stages
  • Live counts and status rings per column
  • Same data as the list & inbox — one source of truth
127.0.0.1:8765
Kanban board with pipeline columns

Every recruiter email, threaded to its job

Sync your inbox and each application email becomes a card — grouped by opportunity, not by time. Upcoming interviews sit up top with the date, stage and interviewer. Every card deep-links back to the real thread.

  • Upcoming interviews, with join links
  • Comms grouped by opportunity
  • Read-only — the agent proposes, you confirm
127.0.0.1:8765
Inbox — recruiter emails grouped by opportunity

Tailored CVs & apply packets

Open any role to a slide-over: status, fit rating, notes, and a CV & Apply panel. Render a clean, ATS-safe CV from your master, let your agent reshape it to the posting, then build an apply packet — mapped fields plus drafted answers, ready for the browser.

  • ATS-safe CV + cover letter (HTML / PDF / TXT)
  • AI-tailored to a specific job
  • Apply packet: fields + answers, stops before submit
127.0.0.1:8765
Job detail slide-over with CV and apply panel
Why it's different

Small, local, and yours.

No SaaS, no lock-in, no telemetry. Just files you own and a toolkit that gets out of the way.

Zero install

Pure Python 3 standard library. No pip, no node, no build step. Clone and run.

Agent-native

Ships with an AGENTS.md playbook so Claude Code, Codex or Cursor knows exactly how to drive it.

Local-first & private

Binds to 127.0.0.1 only. Your pipeline, profile and CVs are plain local files — nothing leaves your machine.

Zero-token ATS scan

Pulls live openings straight from Greenhouse, Ashby, Lever, Workable, Recruitee & SmartRecruiters — no LLM calls.

Two UIs, one engine

A polished web dashboard and a keyboard-driven terminal UI over the same NDJSON pipeline.

MIT licensed

Free and open. Fork it, hack it, make it yours — the whole thing is a few readable Python files.

Quickstart

Running in under a minute.

Clone it and serve. It seeds a small demo dataset so the dashboard is alive immediately — then open the folder in your agent and say “set me up” to make it yours.

Prefer to do it by hand? jobsdb.py setup scaffolds your config from the examples. Run doctor any time to check things over.

Python 3.9+ Chrome (optional, for PDF) Your coding agent
huntbench — zsh
# 1 · clone & serve (demo data included)
$ git clone …/huntbench && cd huntbench
$ python3 jobsdb.py serve
  → dashboard at http://127.0.0.1:8765

# 2 · open in your agent, then say "set me up"
#   it writes config/profile.yml + master-cv.md

# 3 · make it real
$ python3 jobsdb.py reset --yes  # clear demo
$ python3 jobsdb.py scan        # pull live jobs
$ python3 jobsdb.py serve       # triage them

Take command of your job hunt.

Free, open source, and local. Clone it, point your agent at it, and run your search like a pro.

View on GitHub Read the docs →