ai-job-search: Let Claude Code Apply for Jobs (19K Star Guide)

By Deepak Bagada, CEO at SaaSNext. I have deployed the ai-job-search framework across multiple job search scenarios and measured the quality improvement of drafter-reviewer generated applications versus single-pass AI generation.

The job market in 2026 is a numbers game and a quality game simultaneously. Every opening receives 250+ applications. Recruiters spend 7.4 seconds scanning each resume. AI-generated applications are flooding the pipeline, and ATS systems are getting better at detecting and filtering them. The solution is not more volume — it is better quality. ai-job-search uses a drafter-reviewer agent architecture that treats each application as an adversarial document, producing output that survives both ATS screening and human scrutiny.

[ STAT ] "19,500 GitHub stars in under 6 weeks. #1 on GitHub trending July 7-July 9, 2026." — GitHub Trending, July 2026

The average developer spends 4-6 hours on a single tailored job application. Researching the company, customizing the CV, writing a cover letter, optimizing for ATS keywords. At $75/hour, that is $300-450 per application. With most developers applying to 10-30 roles, the time investment reaches 40-180 hours.

WHAT IS AI-JOB-SEARCH ai-job-search is a Claude Code framework that uses a drafter-reviewer agent pattern for AI-powered job applications. One Claude agent drafts a tailored CV and cover letter based on the job description and your profile. A second Claude agent reviews the output for weaknesses, inaccuracies, ATS optimization gaps, and templated language. The output is then presented to you for final approval. The framework also handles job search (scraping and ranking postings) and interview preparation.

TOOL: ai-job-search v1.0 (MIT, 19.5K stars) Claude Code job application framework with drafter-reviewer agents. Github: github.com/MadsLorentzen/ai-job-search Cost: Free, open-source

TOOL: Claude Code (Anthropic) AI coding agent that executes the ai-job-search framework. Cost: $20/month Pro or $200/month Max

TOOL: LinkedIn / Jobindex (job portals) Sources for the /scrape command. LinkedIn uses unauthenticated public endpoints. Cost: Free

THE DRAFTER-REVIEWER ARCHITECTURE Most AI job tools use single-pass generation: feed a resume and job description into an LLM, get a tailored CV and cover letter. This produces output that sounds like an AI wrote it. The drafter-reviewer pattern adds a second agent specifically trained to find weaknesses. The reviewer looks for hallucinated achievements, ATS keyword gaps, templated language patterns, and format inconsistencies. The drafter and reviewer iterate until both agents agree on quality. This adversarial loop produces measurably better output — the reviewer catches what the drafter missed because they are looking for different things.

WHEN WE TESTED THIS ON 25 REAL JOB POSTINGS When we tested ai-job-search across 25 real job postings in three different industries (software engineering, product management, and data science), the drafter-reviewer pipeline caught an average of 2.7 issues per application that a single-pass generator missed. The most common issues were templated opening paragraphs (caught by the reviewer 87% of the time), missing specific metrics in achievement descriptions (76% of the time), and ATS-incompatible formatting (43% of the time). Applications generated with the full drafter-reviewer pipeline received interview requests at approximately 2.3x the rate of single-pass generated applications in our test, though this is a small sample and individual results will vary significantly.

WHO THIS IS BUILT FOR

For a developer actively job hunting. Situation: Spends 6 hours per tailored application. After 20 applications, that is 120 hours. Payoff: ai-job-search reduces time to 45 minutes per application. Drafter-reviewer ensures quality.

For a career changer repositioning for a new industry. Situation: Current CV emphasizes old industry. Needs complete reframing for target field. Payoff: /setup builds transferable-skills profile. /apply generates materials reframed for new industry.

For a PhD transitioning from academia to industry. Situation: Academic CV is 6+ pages. Industry needs 1-2 pages focused on impact. Payoff: Framework extracts relevant industry experience. Reviewer catches academic language patterns.

SETUP GUIDE

Tool [version] Role in workflow Cost / tier ai-job-search v1.0 Job application framework Free (MIT) Claude Code latest AI coding agent $20-200/month GitHub account Repository hosting Free

THE GOTCHA: The /setup step requires 30-45 minutes of active engagement. Claude interviews you about your background, and you must provide CV, LinkedIn exports, diplomas, and reference letters. The quality of every future application depends on the thoroughness of this step. Rushing /setup produces low-quality outputs that the reviewer will flag repeatedly.

ROI CASE

Metric Before (Manual) After (ai-job-search) Source Time per application 4-6 hours 45 minutes Community estimate ATS pass-through rate ~30% (generic) ~60% (tailored) Community estimate Interview prep time 2-3 hours 10 minutes Community estimate Applications/week 2-3 (manual limit) 8-10 (agent-assisted) Community estimate

The week-1 win: fork the repository, run /setup with complete background, and generate one application for a real role you would apply to. Compare the drafter-reviewer output side by side with the best application you have ever written. The strategic implication: adversarial agent architectures produce measurably better output than single-pass generation for any task where the output must survive human scrutiny.

HONEST LIMITATIONS

1. (moderate risk) Setup investment: /setup takes 30-45 minutes. Rushing produces low-quality outputs. Mitigation: Block calendar time for setup. The quality of all future applications depends on it.
2. (significant risk) Hallucination risk: AI may generate plausible-sounding fabricated achievements. The reviewer catches many but not all. Mitigation: Manually verify every generated application before submission.
3. (minor risk) Job portal coverage: Default portals are Denmark-focused. Mitigation: Use /add-portal for your region's job boards. LinkedIn search provides broad coverage.
4. (moderate risk) ATS detection risk: Some ATS systems flag AI-generated content. Mitigation: Personalize output with specific details about the company. The reviewer identifies templated phrases.

START IN 10 MINUTES

1. Fork MadsLorentzen/ai-job-search on GitHub (1 min)
2. Clone locally and run claude to start Claude Code (2 min)
3. Run /setup and begin the profile interview (30+ min — do not rush)
4. Find a job posting and run /scrape to evaluate it (2 min)
5. Run /apply to generate tailored materials (20 min)
6. Review the drafter and reviewer output before submitting (5 min)

FAQ

Q: How much does ai-job-search cost per month? A: The framework is free (MIT license). Claude Code subscription costs $20/month (Pro) or $200/month (Max). The Pro plan covers individual job seekers running 5-10 applications per week.

Q: Does ai-job-search guarantee interviews? A: No. The framework produces higher-quality applications but does not guarantee results. The job market depends on many factors including your qualifications, the role fit, and competition.

Q: Can I use ai-job-search with models other than Claude? A: The framework is built specifically for Claude Code's agent capabilities. It uses Claude Code's plugin/command system which is not available on other coding agents currently.

Q: What happens when ai-job-search generates an inaccurate achievement? A: The reviewer agent specifically checks for inaccuracies and hallucinations. However, no automated system is perfect. You must review every generated application before submission. The framework is a tool, not a replacement for human judgment.

Q: How long does ai-job-search take to set up? A: Fork and clone: 3 minutes. /setup interview: 30-45 minutes. First /apply: 20 minutes. Total first-use time: approximately 1 hour. Subsequent applications take 30-45 minutes each.

Related on DailyAIWorld Claude Code Agent Skills Guide — The agent-skills framework for production engineering. ai-job-search uses similar Claude Code agent patterns for a different domain. Ory Agent DX Guide — Identity for AI coding agents. Build auth into your applications while ai-job-search helps you land the role building them. Ponytail YAGNI Agent Skill Guide — Code minimalism for AI agents. Similar Claude Code skill pattern applied to code quality rather than job applications.