Pick a job. See both readings.
Candidate: about 1.5 years of AI work, Hyderabad, not relocating. The cases mirror an example in the project’s tests; the numbers are illustrative.
JobHunterX reads your resume, finds real openings on companies’ own job boards, and tells you how well each one fits — and how likely a person is to actually read your application. Then it writes honest documents and helps you apply, while you watch.
A public listing on a big job site can collect hundreds of applicants within days. Recruiters read the first few dozen and filter on years, current company and notice period. Early in your career, you’re often filtered out unread. So “apply to more” doesn’t work.
Upload your resume once. It builds a profile — roles, skills and where you used them, location — and shows you a “how we see you” screen to correct before any search.
An employer job board is the careers page a company runs itself, usually through a hiring system like Greenhouse, Lever or Ashby. Posts there are first-hand, and often appear before the big sites copy them. Web search results are only leads — every posting is checked: real, reachable, open, in your country.
Hard checks come first: career track, years, location, must-have skills, notice period. A failed hard check caps the score, so keywords can’t lift a job you can’t get. The ones left get two separate readings — fit, and your chance to be seen.
For the job you pick: a one-page resume, a cover letter under 300 words, and a CV of your whole career. Every AI-written line is fact-checked against your profile. A new number, tool or employer is rejected — your own words stay. Skills you lack are listed for you, never added.
Press Auto-apply and a private browser opens, streamed inside the app, every step in plain words. Stop halts it at once. Take over hands the page to you — for a CAPTCHA, a login, anything. Give back, and it carries on from where you left it.
Illustration on a pretend pageSaved, Preparing, Applied, Interviewing, Offer, Closed. One click to the next stage. Log how you applied and what happened, and see which routes actually get you interviews.
Two small simulations with example data. Nothing here talks to a server, and nothing is submitted anywhere.
Candidate: about 1.5 years of AI work, Hyderabad, not relocating. The cases mirror an example in the project’s tests; the numbers are illustrative.
The agent will hit a “not a robot” check and wait for you.
Each step names the part of the code that does it.
Keep scrolling →Career tracks, realistic titles, skills with evidence. Years computed from dates. Every AI claim is checked against your profile.
intelligence/candidate.pyVaried searches from your own titles and places, aimed at employers’ hiring systems.
discovery/search.pySearch results are leads. Employer boards and watchlist companies are read through public APIs. Out-of-country jobs drop here.
discovery/ats.pyEvery lead becomes one posting: hiring-system API first, then page data, then page text.
discovery/page.pyThe same job from many places becomes one. The first-hand source wins.
discovery/dedupe.pyReal, reachable, current, open. Each field is verified, inferred or unknown — and unknown stays unknown.
discovery/validate.pyRequired vs nice-to-have skills, years, notice, salary. Numbers only count with a quote from the post.
intelligence/job.pyHard checks first. A failed one caps the score. Then weighted parts: role, skills by evidence, experience, location.
intelligence/matching.pyA 0–100 score and verdict, strengths, gaps, unknowns — plus the separate chance to be seen.
intelligence/reach.pyFor top matches: an email the post itself gives (never guessed), search links to recruiters and alumni, a short note in your voice.
intelligence/connect.pyFreshness, channel, applicant counts, direct email routes — each signal listed with its points, next to the fit score and never mixed into it.
About 130 companies in Hyderabad, Bengaluru and remote India: real AI work, early-career hiring, ratings, red flags. Checked every 4 hours while the app is open.
Strong, good, stretch, weak or incompatible — with what fits, what’s missing, what’s unknown, and the exact reason a job was rejected.
For a matched job or any job you paste. Five ATS-safe styles. An AI editor that proposes small fact-checked edits you keep or undo.
Streams into the app with plain-language steps. Handles hidden checkboxes, college vs university, and phone numbers without a double +91.
CAPTCHA or login? Take over, click and type in the live view, press Continue. Progress is saved, even across a restart.
Six stages and a Results view: interviews per 10 applications, by route.
Bars move with real counts the server reports — “Company job boards: 21 of 33” — never estimated from time.
Experience bands instead of a hard cap, a safety-net tab for fresher programmes, a scam filter for fees, UPI and weeks-old websites.
Runs the searches and reads company job boards.
Stamps a posting as real and open — or bins it.
Reads each post and scores the fit.
Finds a way to a real person and drafts a short note.
In the app, Discover → Agents shows nine agents as characters in a 3D office. Every work scene comes from a real event of your running search; break time never claims progress.
Install it like any app. Everything runs on your computer, and new versions offer to install themselves.
JobHunterX_windows_x64-setup.exe.kvcops/JobHunterX release: More info → Run anyway.JobHunterX_darwin_aarch64.dmg.$ chmod +x JobHunterX_linux_amd64.AppImage
$ ./JobHunterX_linux_amd64.AppImageOr install the .deb / .rpm with your package manager and open JobHunterX from your app menu.
Buttons always fetch the newest release · Release notes ↗ · All versions ↗
Never used a terminal? Fine. Copy, paste, press Enter. A script does the rest.
Press Start, type PowerShell, open it.
Press ⌘ Space, type Terminal, press Enter.
Open your terminal — often Ctrl Alt T.
PS> git clone https://github.com/kvcops/jobhunterx.git
PS> cd jobhunterxPS> powershell -ExecutionPolicy Bypass -File start.ps1Or double-click start.bat in the folder. The bypass applies to this one run only.
$ bash start.shIt installs uv, makes a private environment, installs packages (about 30 seconds), creates the settings file, checks everything loads, and opens the app in your browser.
The first screen shows where to get a free key, tests it and saves it. Google AI Studio is easiest; NVIDIA, Groq and Mistral work too. No key? Start free uses Kilo’s free models — marked as possibly training on your prompts, which include your resume text.
The app lives at http://127.0.0.1:9000 — on your own computer. Stop: Ctrl + C. Next time, run the same start command.
PS> powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"$ curl -LsSf https://astral.sh/uv/install.sh | shPS> uv venv --python 3.11
PS> .venv\Scripts\Activate.ps1$ uv venv --python 3.11
$ source .venv/bin/activatePS> uv pip install -r requirements.txt
PS> cd jobhunterx
PS> python -m jobhunterx.api.main| Resume upload keeps loading | No real AI key. Restart; the first screen asks for one. |
|---|---|
uv not found | Reopen the terminal, or pip install uv. |
No module named jobhunterx | Run from the inner jobhunterx folder with the venv on. |
| Port 9000 busy | Set PORT=9001 in jobhunterx/.env, open http://127.0.0.1:9001. |
| Auto-apply can’t open a browser | Install Google Chrome. |
| Searches are slow | Add a second free key. Free plans allow few calls a minute. |
On your computer — the desktop app’s own folder, or a local database from source. No JobHunterX account, no JobHunterX cloud.
Resume and job-post text go to the AI provider you choose. Kilo’s free models may train on prompts — add any free key and turn Kilo off in Settings if that matters.
Free AI plans allow few calls a minute. CAPTCHAs and logins may need you. Installers aren’t signed yet. No tool can promise interviews.
Job seekers in India. The watchlist covers Hyderabad, Bengaluru and remote India today. Sharpest for 0–3 years of experience.
An app on your computer for the whole job search: it understands your resume and goals, finds real openings, explains fit and chance to be seen, writes honest documents, fills applications with a browser agent you can take over, and tracks everything.
Yes and yes — MIT licensed, all of it on GitHub. It uses AI providers’ free tiers; their limits can change.
No. Searching runs on its own; Auto-apply starts when you press it on a job you chose. You watch live and can stop, take over, or finish and mark it submitted yourself.
The application shows “needs you”. Take over, click and type in the live view, press Continue. Progress is saved.
Not strictly — “Start free” works with no key. A free Google AI Studio key is the easy private option.
Installers: Windows 10/11 x64, macOS 11+ Apple Silicon, Linux x86-64 (AppImage, .deb, .rpm). Other machines can run from source.
Alerts forward keywords. JobHunterX checks a post is real and open, reads it, applies hard checks, explains the score, estimates your chance of being seen, and goes to employers’ own boards first.
No. It helps you send fewer, better-fitted applications, earlier, with honest documents.