CareerRat / Terminal workspace

Your job hunt.
In your terminal.

CareerRat turns the AI CLI you already have into a personal recruiter. It rates the jobs worth chasing, applies with your real experience, and tracks what happens. Your saved jobs on one side. Your agent on the other.

Start here

npm i -g careerrat
careerrat start

Bring Claude Code or Codex.
Your workspace stays on your machine.

MIT licensedLocal SQLite
The CareerRat terminal workspace: a table of saved jobs with fit and pay on the left, details for the selected job below it, and the agent chat on the right.
The terminal workspace, shown with demo jobs. Run careerrat start to open yours.

01 / How it works

Rate. Apply. Track.

01

Open it

Run careerrat start. It finds the AI CLI on your machine and sets itself up. No account, no forms.

LOCAL AI CLI · READY
02

Tell it what you want

Drop your résumé, then answer short questions in plain English, like what would make one job worth applying to before another.

ROLES ✓ · GUARDRAILS ✓
03

It hunts, you decide

Search opens when setup is ready. It looks in built-in job boards, and finds specialist boards and employer pages on its own as it learns. Nothing counts until Evaluate reads the full posting. You review any missing answers before it fills anything, and CAPTCHAs stay with you. CareerRat submits only after its checks pass; choose manual mode to press Submit yourself.

SWEEP · 9 CUT · 12 RANKED
The Fit report tab listing each requirement from a posting with the evidence from your record and a strong, partial, or gap match.
The Fit report checks the posting against your must-haves, one requirement at a time.
The Activity tab showing a search running in the background with sources checked and jobs found so far, above a finished search with a Review new matches button.
Search jobs runs in the background. Keep working while Activity shows its progress.

What stays in your hands

  • The terminal search worker skips browser-backed saved searches and does not open a login window for them. Ask the agent to research login-gated sites separately.
  • It fills safe application fields for you, but only after you review any missing answers.
  • Voluntary questions stay blank by default; local Application defaults can choose a decline option when available.
  • Anything found this way stays unverified until Evaluate reads the full posting for location, office days, pay, and fit.

02 / Your side of the table

An agent for your side of the table.

Companies screen you with software. CareerRat aligns your experience to the role, gets you past the ATS, and only spends your time on jobs you could actually get.

Rate before you apply

Every full posting gets checked against your location, pay, and other must-haves before you write a word. No cover letters for jobs that don't fit.

Aligned, not fabricated

It shapes your real experience to fit the role, and it never invents anything that isn't yours.

Search broadly, verify honestly

Built-in job-board sources search for the roles you set. CareerRat also searches the open web to catch postings other tools miss, and finds specialist boards and employer pages as it learns your search. Every open-web lead stays unverified until Evaluate actually reads the posting.

On your machine

Your workspace lives on your machine. Only the context a task needs goes to the AI provider you picked, under its terms. Searches keep running in the background while you move between tabs, and if something gets interrupted, you get a clear retry instead of losing your place.

03 / Get CareerRat

One install, and you're talking.

Run npm i -g careerrat, then careerrat start. The workspace opens in your terminal with your agent beside your jobs, and you say hi. Setup is a conversation, not a form. Open source under MIT.

Install with npm

Needs Node.js 24.22.1 or newer. Also needs Python 3.11 or newer; the first run takes about a minute to set up.

You'll need an AI CLI

Pick Claude Code (2.1.241 or newer) or OpenAI Codex. Either one runs the same CareerRat-owned workflows and skills. CareerRat invokes it directly and never falls back to another provider.

Automatic picks the right level for each task: complex judgment stays strong, while web searches and small helpers stay efficient. You can instead choose Faster, Balanced, or Best, and set Thinking depth to Automatic, Low, Medium, or High. Those controls work the same with Claude Code or OpenAI Codex.

  • Claude Code
  • OpenAI Codex
  • Same CareerRat workflow

A runtime appears ready only when it is available, signed in, and passes its readiness check. CareerRat itself costs nothing; your AI provider may have its own plan or usage costs. Update any time with careerrat update.

04 / Open source

Star history

Star history chart for CodesWhat/careerrat

Ready when you are.

Install it and start talking.