# career-ops and ai-job-search Without a Terminal: The Same AI Job Search Inside the Claude App (2026)
> career-ops (73k stars) and ai-job-search (44k stars) are the best free AI job-search systems on GitHub, and both run in a coding terminal. Here is what they do, who they suit, and a free no-terminal alternative that runs inside the Claude app with Gmail, Calendar and Drive.
**Author:** [Alex Lowe](https://theaicareerlab.com/about) — Founder, The AI Career Lab
**Published:** 2026-09-27
**Canonical URL:** https://theaicareerlab.com/blog/career-ops-ai-job-search-without-terminal
**Category:** comparison
**Tags:** job search, career-ops, ai-job-search, Claude, resume, 2026
---> **TL;DR.** [career-ops](https://github.com/career-ops-hq/career-ops) and [ai-job-search](https://github.com/MadsLorentzen/ai-job-search) are the two best-known open-source AI job-search systems of 2026, with roughly 73,000 and 44,000 GitHub stars between them as of September 27. Both are genuinely good, both are free, and both run in a coding terminal: you fork or install a repo, edit config files, and drive the workflow from Claude Code or a similar CLI. If that describes you, use them. If it does not, the **Job Search AI Operating System** runs the same workflow (evaluate the posting, tailor the resume, track the pipeline, prep the interview, negotiate the offer) inside the Claude app as a plugin, with drafts that land in your own Gmail, Calendar and Drive behind an approval step. It is free too. The honest trade-off: it needs a paid Claude plan, and it does not scrape job boards.

| If you… | Use | Because |
|---|---|---|
| Live in a terminal and want the widest net | **career-ops** | 100-plus pre-configured companies, 55-plus job-board modules, A-to-H scoring, runs on several AI CLIs including a free tier |
| Want a polished, verified LaTeX application pipeline on Claude Code | **ai-job-search** | Drafter-reviewer workflow, page-by-page PDF checks, ATS check on the text layer |
| Have never opened a terminal and already use Claude | **Job Search AI Operating System** | Plugin install in a few clicks, five connected jobs, drafts into Gmail, Calendar and Drive with approval, profession archetypes for non-tech roles |
| Want scraping plus a no-terminal daily driver | **Both** | Scan with career-ops, then run applications, inbox, prep and negotiation in the Claude app |

## What the two repos actually do

Both projects were built by someone running their own search, and it shows: the workflows are opinionated in useful ways.

**ai-job-search** was written by a geophysicist whose position was cut in late 2025. The README describes the outcome plainly: 69 tailored applications, 20 first interviews, one signed contract, and a new role as an AI engineer in June 2026. The framework turns Claude Code into an application assistant with three core commands. `/setup` builds a profile from your documents, a pasted CV or an interview. `/scrape` searches job portals (the shipped ones are Danish: Jobindex, Jobnet, Akademikernes Jobbank, plus LinkedIn and a multi-market aggregator) and ranks matches by fit. `/apply` runs a drafter-reviewer pipeline: evaluate fit, draft a LaTeX CV and cover letter, spawn a second agent to research the company and critique the drafts, revise, then compile the PDFs and inspect them until the CV is exactly two pages with no orphaned titles. It then extracts the PDF's text layer the way an applicant tracking system would and checks keyword coverage, with an explicit rule that a keyword your profile does not support is recorded as a gap, never stuffed in. Ten more commands cover interview prep from the exact materials the interviewer read, outcome tracking, a Notion pipeline view, an optional Gmail sync that proposes status changes for you to approve, a skills-gap heatmap, and an offline HTML dashboard.

**career-ops** is broader and more portable. You bootstrap a project folder with `npx @santifer/career-ops init`, add a markdown CV and a YAML profile, and drive it from Claude Code, Codex, OpenCode, Antigravity, Grok, Kimi or Qwen. Paste a job URL and it runs the auto-pipeline: an archetype detection step, an A-to-H evaluation (role summary, weighted CV match, level strategy, compensation research, personalisation, STAR-plus-reflection interview prep, a posting-legitimacy check for ghost jobs and scams, and a work-authorisation flag) that ends in a global score from 1 to 5, then an ATS-optimised PDF, a cover letter draft and a tracker entry. It will tell you not to apply below 4.0 and say so if you override it. The scanner ships with more than 100 companies and 55-plus provider modules across Ashby, Greenhouse, Lever, Wellfound and feeds. Around that core sit more than 40 modes: contact discovery with a LinkedIn note under 300 characters, deep company research, a story bank, negotiation scripts, follow-up cadence, reply classification, rejection-pattern analysis, and a calibration report that checks whether your evaluation scores predict your real outcomes.

Both are MIT-licensed, both were updated the day this was written, and both are explicit about what they will not do. career-ops will not auto-submit an application, will not send an email (there is no mail transport in the codebase) and will not phone home. ai-job-search treats postings as untrusted input, drafts follow-ups but never sends them, and caps follow-ups at two per application.

## Who they are for, and who they are not

Read the prerequisites and the audience is clear.

ai-job-search asks for Claude Code, Python 3.10 or later, Bun for the portal tools, and a LaTeX distribution with `lualatex` and `xelatex`, with a note that minimal TeX installs need extra packages. Your profile lives in tracked files in a git repository, so the README warns you to use a private repo rather than a public fork. career-ops asks for Node, Playwright (with a note about Chromium's system libraries on non-Debian Linux), a `config/profile.yml`, a `portals.yml` and a `cv.md`, and its own `npm run doctor` to validate the setup. Its default archetypes are LLMOps, agentic engineering, product management, solutions architecture, forward-deployed engineering and transformation, and it invites you to ask your CLI to change them.

None of that is a criticism. It is the natural shape of a tool built by and for people who already work in a terminal. It is also why, for a nurse, a paralegal, a loan officer, a teacher or an operations manager, these repos mostly stay unopened. The workflow they encode is universal. The packaging is not.

The other limit is action. Both projects draft; neither one touches your calendar or files a document in your Drive, and career-ops in particular treats "never sends" as a design principle. For a developer that is the right call. For someone who will be living in Gmail and Google Calendar for the next three months, it means a lot of copy and paste.

## The same workflow, inside the Claude app

The [Job Search AI Operating System](https://clowealex.gumroad.com/l/job-search-ai-os) was built for the second group. It is a Claude plugin, and the entire setup is: in the Claude app, open Customize, then Plugins, choose Add marketplace, then Add from a repository, paste `alexclowe/job-search-os`, and click Add. (If you prefer a zip, the free kit includes one you upload from the same menu.) Then say "set up my Job Search OS" and paste your resume; setup takes about two minutes and stores your career profile in a Project so every later conversation finds it.

From there you work from a command center, the way the profession plugins on this site do. It reads your trackers, shows the state of your search in one line (active applications, interviews this week, offers in play, follow-ups due), and offers five jobs:

1. **Tailor and apply.** A posting plus your resume (no resume? it interviews you and builds a master resume first). Before writing a word it shows one check card: whether the posting shows the caution signals official guidance names for job scams, the pay against your minimum (hourly against hourly, salary against salary), how well you fit, and whether you know anyone there, with a choice to tailor it, skip it, or ask for the intro first. Then it produces a tailored resume and cover letter that lead with what you did and what came of it rather than what you were called, lists the posting's terms your resume now uses and asks about the ones your record can't support (it never adds them on its own), drafts the application-form answers, files everything in a per-company Drive folder, and adds the row to your Application Board.
2. **Work my inbox.** Reads recruiter and hiring-team threads you have confirmed, drafts replies in your voice inside the thread, puts interviews on your calendar with a prep hold, and updates the board. It can run on a schedule.
3. **Prep for an interview.** Starts with a one-page company brief where every fact has a source and a date, then builds a stage-specific brief from the round and your Story Bank: likely questions, your stories mapped to their concerns, and the questions to ask back. Then it offers a practice round, one question at a time, in the format your profession's interviews usually take.
4. **Follow up and negotiate.** Thank-you notes, a follow-up cadence with dates on the calendar, and a counter built on your minimum pay, logged on an Offer Tracker with a spreadsheet that does the arithmetic.
5. **Run my weekly review.** What moved, what went stale, your reply rate by source and the stage where applications stall (it says "too few to tell yet" under about ten), three moves for the week, outreach drafted, and focus blocks on your calendar.

Five guards run inside every job. The "is this real?" check lists scam and not-an-open-role signals as signals, never verdicts. The salary guard flags any posting, message or offer below your number before materials are written and records "applying anyway" if you proceed. The oversell guard makes every claim survive the question "what exactly did you do?", downgrades verbs to match the record, and strips the AI-register words that recruiters and detection tools have learned to spot. The fabrication guard traces every metric, title, date and employer back to your resume, profile or Story Bank, marks anything else as unconfirmed, and never rounds up. And the undersell check does the reverse of the oversell guard against the same record, flagging a real win the draft buried or left out. A people tracker reads your LinkedIn connections export and keeps only the people at companies you're applying to, so "ask for the intro first" is one click.

It does not assume you are an engineer, or that you're paid a salary. The profile asks for your target roles, your career stage (from first job to returning to work), how you're paid (hourly, salary, commission, or a salary schedule) and the lowest you'd take in that unit, anything a hiring manager might question, and three to five things you did and what came of them; nothing assumes an engineering resume or a LaTeX template. Where career-ops ships six archetypes for technical roles, this one ships a profession archetype registry (nurse, teacher, bookkeeper, loan officer, paralegal, real estate agent, social media manager, personal trainer to start, with a template for adding more), and every job reads the archetype for your profession: target titles, where the jobs are posted, what recruiters screen for, the story types that land, how interviews usually run, how pay is structured, and what's usually negotiable. And because the plugin runs where the rest of your Claude work lives, every draft goes into your own Gmail, Calendar and Drive, and every send waits for you to approve it. Nothing accepts, declines or resigns on your behalf, and the negotiation scripts never volunteer your salary history.

There are two first-run samples, a nurse paid hourly and a data engineer paid a salary, so you can watch a full Tailor and apply pass on a fictional candidate before you paste anything of your own.

## What it does not do, and what it costs

Two honest gaps. It does not scrape job boards; you bring the posting. career-ops's scanner is genuinely valuable if you are running a wide search across many companies, and nothing in this plugin replaces it. And it does not compile LaTeX; resumes and letters are Google Docs in your Drive, which is what most non-technical hiring processes want anyway.

On cost: the plugin and the kit are free (pay what you want, starting at $0). Claude plugins, though, are not available on Claude's Free tier, so you need Pro or above. career-ops can run at zero cost on a free CLI tier and cheaply on local models, per its own guides; ai-job-search's README is upfront that Claude Code needs a paid plan or API credits. If you are already paying for Claude, the plugin costs nothing extra. If you are not, the repos may be the cheaper path, provided you are comfortable installing them.

## Watch it work

Six short narrated recordings of the plugin running on claude.ai against a fictional sample search. Nothing in them is sent, submitted, or accepted without a click.

- [First run: install, setup, one sample application (2:31)](https://images.theaicareerlab.com/video/job-search-os-first-run.mp4)
- [Setup: resume to command center (1:25)](https://images.theaicareerlab.com/video/job-search-os-setup.mp4)
- [Tailor & apply, salary check first (1:27)](https://images.theaicareerlab.com/video/job-search-os-tailor-apply.mp4)
- [Prep for an interview from the Story Bank (1:27)](https://images.theaicareerlab.com/video/job-search-os-interview-prep.mp4)
- [Follow up & negotiate with the offer worksheet (1:27)](https://images.theaicareerlab.com/video/job-search-os-negotiate.mp4)
- [Run my weekly review (1:16)](https://images.theaicareerlab.com/video/job-search-os-weekly-review.mp4)

## Which one should you pick?

- **You are a developer or a PM, you already use Claude Code or Codex, and you want the widest possible net.** Start with career-ops. Run the scanner, let the A-to-H score triage the pile, and use its story bank and negotiation modes.
- **You want the most careful application artifact you can produce, and you like LaTeX.** ai-job-search's drafter-reviewer loop and PDF verification are the strongest application pipeline of the three.
- **You have never opened a terminal, or you would rather not, and you already pay for Claude.** Install the Job Search AI Operating System. Run the sample, then Tailor and apply on a real posting.
- **You want both breadth and a daily driver.** Scan with career-ops, then paste the postings it surfaces into Tailor and apply and let the plugin run your inbox, prep, follow-ups and weekly review from the Claude app.

If you only need the resume step today, the [five Claude prompts to tailor a resume to a job description](/resources/claude-resume-optimization-prompts) work in any AI, and the [Resume Optimizer](/tools/resume-optimizer) does it in one pass. For the wider picture on where the jobs are moving, start at the [careers hub](/careers).

## Sources

1. career-ops repository (README, features, "What career-ops will not do", tech stack, portals): [https://github.com/career-ops-hq/career-ops](https://github.com/career-ops-hq/career-ops) (star and fork counts read from the GitHub API on September 27, 2026)
2. ai-job-search repository (README, prerequisites, "How /apply works", other commands): [https://github.com/MadsLorentzen/ai-job-search](https://github.com/MadsLorentzen/ai-job-search) (star and fork counts read from the GitHub API on September 27, 2026)
3. career-ops free-tier and budget guides: [https://github.com/career-ops-hq/career-ops/blob/main/docs/FREE_TIER.md](https://github.com/career-ops-hq/career-ops/blob/main/docs/FREE_TIER.md) and [https://github.com/career-ops-hq/career-ops/blob/main/docs/RUNNING_ON_A_BUDGET.md](https://github.com/career-ops-hq/career-ops/blob/main/docs/RUNNING_ON_A_BUDGET.md)
4. ai-job-search community forks and adaptations: [https://github.com/MadsLorentzen/ai-job-search/discussions/78](https://github.com/MadsLorentzen/ai-job-search/discussions/78)
5. Claude plans and plugin availability: [https://claude.com/pricing](https://claude.com/pricing)
6. Job Search AI Operating System (free kit): [https://clowealex.gumroad.com/l/job-search-ai-os](https://clowealex.gumroad.com/l/job-search-ai-os) and repository: [https://github.com/alexclowe/job-search-os](https://github.com/alexclowe/job-search-os)
## Frequently asked questions

### Is there a career-ops alternative that doesn't need a terminal?

Yes. The Job Search AI Operating System from The AI Career Lab runs the same evaluate-tailor-track-prepare-negotiate workflow inside the Claude app (web, desktop and mobile) as a plugin. You install it from a GitHub marketplace in a few clicks or upload a zip; there is no git, Node, Python, LaTeX or YAML. It is free (pay what you want, $0 minimum). It does need a paid Claude plan, because Claude plugins are not available on the Free tier.

### Do career-ops or ai-job-search send emails or submit applications for me?

No. Both are deliberately draft-only. career-ops states it will not auto-submit an application, will not send an email, and has no mail transport in its codebase. ai-job-search's follow-up and thank-you drafts are drafts only, never sent, and its optional Gmail sync only reads status signals and asks you to approve any tracker change. The Job Search AI Operating System drafts into your own Gmail, Calendar and Drive, but every send still waits for your approval.

### Which is better, career-ops or ai-job-search?

They overlap heavily and both are excellent. career-ops has the broader portal scanner (100-plus pre-configured companies, 55-plus job-board modules), a structured A-to-H evaluation with a 1-to-5 score, rejection-pattern analysis and support for several AI CLIs. ai-job-search is built tightly around Claude Code with a drafter-reviewer application pipeline, LaTeX CVs verified page by page, and an ATS check on the PDF's text layer. If you are comfortable in a terminal, try career-ops first for breadth; pick ai-job-search if you want the polished LaTeX application flow.

### Can I run career-ops or ai-job-search for free?

career-ops documents a zero-cost path on Antigravity CLI's free tier and a budget guide for local or cheaper models. ai-job-search is built on Claude Code, which its README notes has no free tier (you need a Claude Pro, Max or Team subscription or API credits); community forks adapt it to Codex, Antigravity and Gemini CLI. The Job Search AI Operating System is itself free, but plugins require a paid Claude plan (Pro and up).

### Is the Job Search AI Operating System open source?

The plugin is published in a public GitHub repository, alexclowe/job-search-os, which doubles as a one-plugin Claude marketplace. The free kit on Gumroad adds the printable guides. Use whichever install path you prefer.

---

*Canonical version: https://theaicareerlab.com/blog/career-ops-ai-job-search-without-terminal*
*This document is the Markdown companion served for AI crawlers and answer engines. See the canonical URL for the rendered version with navigation, related content, and interactive elements.*