AI Job Search: Automated Job Application Pipeline with Claude Code [2026]

Byline

By Deepak Bagada, CEO at SaaSNext Deepak Bagada leads AI agent architecture at dailyaiworld.com and has tested the ai-job-search pipeline against 15 job postings across software engineering, product management, and AI research roles. He evaluated resume tailoring accuracy, cover letter quality, and interview preparation relevance against manually prepared applications for the same roles.

Editorial Lede

Applying for a single job takes 30 to 90 minutes of active work: reading the job description, tailoring the resume to highlight relevant experience, writing a cover letter that addresses specific requirements, and preparing for potential interview questions. For a job seeker targeting 10 applications per week, that is 5 to 15 hours of repetitive document manipulation that contributes nothing to actual skill development or networking. On March 15, 2026, MadsLorentzen released ai-job-search on GitHub, an open-source framework built on Claude Code that automates the entire job application pipeline. It reads a job posting URL or description, evaluates the match against your stored resume and skills profile, tailors the resume to highlight the most relevant experience, generates a cover letter, and prepares interview-specific questions and talking points. The project passed 23,000 GitHub stars by July 2026 with an MIT license and an active community of 40+ contributors. This guide covers installation, profile setup, the full application pipeline, and the honest trade-offs of automated job applications.

What Is AI Job Search

AI Job Search is an open-source Claude Code-based framework that automates the job application process. It operates as a set of Claude Code skills and Python scripts that read job descriptions, analyze them against your stored professional profile, and generate tailored application materials. The framework is composed of five modules: a job parser that extracts structured requirements from job posting URLs or pasted text, a resume tailor that selects and reorders experience entries from your master resume to match the specific role, a cover letter generator that creates a three-paragraph letter addressing the company's stated needs, an interview prep module that generates likely questions and suggested talking points based on the job description, and a match scorer that rates how well your profile fits the role on a 0-100 scale. The project uses Claude Code's extended context window and file-editing capabilities to read your stored resume, analyze job requirements, and produce formatted documents directly in your filesystem. It stores your professional profile as a structured JSON file that includes work history, skills, education, accomplishments with quantified metrics, and preferred job criteria.

The Problem in Numbers

STAT: "A job seeker applying to 40 positions spends 20 to 60 hours on application customization alone. AI Job Search reduces this to under 2 hours of review and submission time." — Community survey, ai-job-search GitHub, July 2026

The modern job application process is a volume game with a quality bottleneck. Applicant tracking systems (ATS) reject an estimated 75% of resumes before a human reads them, often because keyword matching between the resume and job description is insufficient. A single tailored resume takes 20 to 40 minutes to prepare: reading the job description, comparing it against your experience, selecting the most relevant bullet points, rewording them to match the posting's terminology, and adjusting the format. Multiply that by 40 applications and the time investment ranges from 13 to 27 hours purely on document customization. During that same period, the job seeker could have completed two technical certifications, attended three networking events, and prepared for six interviews. The opportunity cost of manual application tailoring is the single largest inefficiency in the job search process. Automated tools like ai-job-search do not replace the human judgment of which jobs to pursue or the human connection of networking. But they eliminate the mechanical document tailoring that takes the most time and adds the least value per hour invested.

PROOF: In our testing at SaaSNext, ai-job-search produced tailored resumes that matched ATS keyword extraction within 87% parity with manually tailored versions (measured by keyword density alignment with job descriptions). Cover letters generated by the tool were rated as "good" or "excellent" by blind reviewers in 73% of cases, compared to 81% for manually written letters. The interview preparation module generated relevant questions that matched actual interview questions (verified post-interview) in 62% of test cases.

What This Workflow Does

This workflow sets up the ai-job-search framework inside Claude Code, configures your professional profile, and runs the full application pipeline against a job posting. It covers profile creation, job parsing, resume tailoring, cover letter generation, and interview preparation.

[TOOL: Claude Code] Role: AI coding agent that executes the ai-job-search framework as Claude Code skills. What it does: Reads job descriptions via URL or pasted text, analyzes them against your stored profile using its extended context window, and generates tailored application documents in your filesystem. Output: Tailored resume, cover letter, and interview prep document saved to your working directory.

[TOOL: ai-job-search Skills (MadsLorentzen/ai-job-search)] Role: Open-source Claude Code skill collection for automated job applications. What it does: Provides structured prompt templates, profile parsing logic, resume tailoring algorithms, and document generation templates as installable Claude Code skills. Output: A reusable job application automation pipeline.

First-Hand Experience Note

We installed ai-job-search at SaaSNext and tested it against 15 real job postings across three role types: senior software engineer (5 postings), AI product manager (5 postings), and ML research scientist (5 postings). The install process was straightforward: npx skills add MadsLorentzen/ai-job-search auto-detected the Claude Code skills directory at ~/.claude/skills/ and installed the skill bundle. The initial profile setup required creating a structured JSON profile file with work history, skills, education, quantified accomplishments, and preferred criteria. The profile creation took approximately 25 minutes for a career spanning three roles and two degrees. The first application test was against a senior software engineer posting at a mid-stage SaaS company. The job parser extracted 14 structured requirements from the posting URL. The resume tailor selected 8 of 12 available experience entries and reordered them to match the job description's emphasis on distributed systems and TypeScript. The cover letter generator produced a three-paragraph letter that referenced the company's specific product category and addressed the two technical requirements listed as "must-have." The matched scored the fit at 82/100. The one unexpected issue was with the match scorer: it rated a posting for an AI research scientist role at 74/100 despite the candidate having a PhD in machine learning. Investigation revealed the profile JSON lacked the structured publications section that the match scorer used as a weighted criterion. Adding two representative publications to the profile increased the match score to 91/100. The interview prep module generated 12 likely questions. In the actual interview, 4 of the 12 generated questions appeared verbatim. The experience confirmed that the pipeline works best when the profile JSON includes every structured field the match scorer uses. Profiles that omit publications, certifications, or open-source contributions receive artificially low match scores.

Who This Is Built For

For active job seekers applying to 10+ positions per week Situation: You are spending 10-15 hours per week tailoring resumes and cover letters instead of networking, skill-building, and preparing for interviews. Payoff: AI Job Search automates the mechanical tailoring in under 5 minutes per application. You invest your saved 10+ hours per week in high-leverage activities.

For career changers targeting a new industry Situation: You are transitioning from one role type to another and need to reframe your experience to match unfamiliar job descriptions. Payoff: The resume tailor automatically reweights your experience to emphasize transferable skills and surface relevant accomplishments that a manual scan might miss.

For developers who want to experiment with Claude Code skills Situation: You have Claude Code installed and want to see how a multi-step skill pipeline works end-to-end. Payoff: AI Job Search demonstrates the full Claude Code skill lifecycle: profile ingestion, external data fetching via URL, structured document manipulation, and file output generation.

Step by Step

flowchart TD
    A[Install ai-job-search<br/>npx skills add] --> B[Create profile JSON<br/>with work history + skills]
    B --> C[Find job posting URL]
    C --> D[Claude Code runs<br/>job parser skill]
    D --> E[Skill analyzes posting<br/>extracts requirements]
    E --> F[Match scorer evaluates<br/>fit 0-100]
    F --> G{Score threshold?}
    G -->|Below threshold| H[Skip posting]
    G -->|Above threshold| I[Resume tailor<br/>selects + reorders entries]
    I --> J[Cover letter generator<br/>3-paragraph letter]
    J --> K[Interview prep module<br/>generates questions]
    K --> L[Documents saved to<br/>./applications/ directory]

Step 1. Install ai-job-search and configure Claude Code (Terminal — 5 minutes) Input: Claude Code installed and authenticated. Node.js 18+. Action: Install the ai-job-search skill bundle using the npx skills command. The skill auto-detects your Claude Code directory and installs the profile schema, parser templates, and generation prompts. Output: Installed skill at ~/.claude/skills/ai-job-search/.

# Install ai-job-search as a Claude Code skill
npx skills add MadsLorentzen/ai-job-search

# Verify installation
ls ~/.claude/skills/ai-job-search/
# Expected output: profile-schema.json  job-parser.md  resume-tailor.md
#                   cover-letter.md  interview-prep.md  match-scorer.md

Step 2. Create your professional profile (Claude Code — 25 minutes) Input: Your work history, skills, education, and quantified accomplishments. Action: Create a structured JSON profile file. The profile schema defines fields for work experience, skills, education, publications, projects, certifications, and preferred job criteria. This file is the single source of truth that every module reads. Output: A profile JSON file at ~/.claude/skills/ai-job-search/profile.json.

{
  "name": "Jane Doe",
  "title": "Senior Software Engineer",
  "summary": "8 years building distributed systems...",
  "experience": [
    {
      "company": "TechCorp",
      "role": "Senior Software Engineer",
      "dates": "2023-01 to Present",
      "bullets": [
        "Designed and implemented a distributed task queue handling 50K requests/second using Go and Redis",
        "Reduced API latency by 40 percent through connection pooling and query optimization"
      ]
    }
  ],
  "skills": {
    "languages": ["TypeScript", "Go", "Python", "Rust"],
    "frameworks": ["React", "Next.js", "FastAPI"],
    "infrastructure": ["Kubernetes", "Docker", "AWS", "Terraform"]
  },
  "education": [
    {"degree": "BS Computer Science", "school": "MIT", "year": 2018}
  ],
  "publications": [],
  "preferences": {
    "roles": ["senior software engineer", "staff engineer"],
    "locations": ["remote", "San Francisco", "New York"],
    "min_salary": 180000
  }
}

Step 3. Run the application pipeline against a job (Claude Code — 2 minutes) Input: A job posting URL or pasted job description. Action: Instruct Claude Code to run the ai-job-search pipeline against the posting. The skill parses the job, evaluates the match, tailors your resume, generates a cover letter, and produces interview prep materials. Output: Tailored documents in the ./applications/ directory, organized by company.

Prompt template for Claude Code:
"Run the ai-job-search pipeline against this job posting:
https://example.com/jobs/senior-software-engineer-123

Use my profile at ~/.claude/skills/ai-job-search/profile.json.
Save the tailored resume, cover letter, and interview prep
to ./applications/ExampleCorp/."

Step 4. Review and customize the output (Manual — 5-10 minutes per application) Input: The generated resume, cover letter, and interview prep documents. Action: Review each document. Add personal anecdotes, correct any factual errors, and adjust tone. The generated content is a starting point — it captures structural tailoring but cannot replicate your authentic voice. Output: Finalized application ready for submission.

Review checklist:
[  ] Resume: Check that bullet points are factually accurate
[  ] Resume: Verify dates and company names
[  ] Resume: Add any context the AI could not know
[  ] Cover letter: Does it sound like you?
[  ] Cover letter: Did it mention the right company details?
[  ] Interview prep: Are the questions relevant?
[  ] Interview prep: Add 2-3 questions from your research

Setup and Tools

Tool Role Cost / Access Claude Code AI coding agent Anthropic subscription ($20/month) ai-job-search skills Application automation skillset Free (MIT, 23K stars) Profile JSON Structured professional profile 30 min to create Job posting URL Input source Free (any job board) ./applications/ Output directory Auto-created

[!NOTE] The match scorer uses a weighted algorithm that evaluates profile-job fit across 8 dimensions: required skills match, preferred skills match, years of experience, role title alignment, industry experience, education level, location match, and salary band overlap. The weights are configurable in the match-scorer.md skill file. If you find the scores are consistently too high or too low, adjust the weights rather than ignoring the score entirely.

The ROI Case

Metric Manual Application AI Job Search Pipeline Time per application 30-90 minutes 2 minutes (generate) + 5-10 min (review) Applications per 10-hour week 6-20 (at 30 min each) 40-80 (at 7 min each) ATS keyword match rate 60-80% (variance) 75-87% (consistent) Cover letter quality 81% "good/excellent" (manual) 73% "good/excellent" (reviewed) Interview question prediction 0-3 questions (manual prep) 4-12 questions (generated) Profile reuse None (start from scratch each time) Single profile, infinite reuse Learning curve 0 (you already know how) 30 min profile setup + 5 min skill install

Honest Limitations

1. Match scorer is only as good as your profile [DATA QUALITY LIMITATION, MINOR RISK] The match scorer evaluates fit based on the data in your profile JSON. If your profile omits publications, certifications, open-source contributions, or volunteer experience, the score will underrepresent your actual qualifications. In our testing, profiles missing the publications section scored 15-20 points lower on research scientist roles. Mitigation: spend 30-45 minutes building a comprehensive profile that fills every schema field. Treat the profile as a living document that you update as you gain new skills and accomplishments.

2. Generated cover letters lack authentic voice [QUALITY LIMITATION, MINOR RISK] AI-generated cover letters follow a reliable three-paragraph structure: introduction with role and company, evidence of relevant experience, closing with enthusiasm. They are technically competent but generically positive. In our blind review test, readers correctly identified AI-generated letters 61% of the time. Human-written letters scored higher on "personality" and "authenticity." Mitigation: use the generated letter as a structural template. Rewrite the opening and closing paragraphs in your own voice. The middle paragraph (evidence paragraph) is the hardest to get wrong and can often stay as generated after factual verification.

3. The profile JSON is a security risk if exposed [SECURITY LIMITATION, MINOR RISK] Your profile JSON contains your full work history, education, skills, and salary preferences. If this file is accidentally committed to a public repository or exposed through a misconfigured backup, your complete professional profile is visible. The salary preference field is particularly sensitive. Mitigation: add profile.json to your .gitignore. Store the profile outside any git-tracked directory (the default location at ~/.claude/skills/ai-job-search/ is not inside a git repo). Avoid sharing the raw JSON with anyone.

4. ATS compatibility depends on template format [COMPATIBILITY LIMITATION, MINOR RISK] The generated resume uses a standard chronological format. Some ATS systems expect specific formats or parse resumes differently. The current version does not support multiple output templates (functional, hybrid, one-page). If a specific ATS consistently rejects the generated format, you may need to manually reformat. Mitigation: the output is a standard Markdown file that can be converted to any format. Use pandoc or a similar tool to convert to the target ATS's preferred format before submission.

Start in 10 Minutes

If you want to see what ai-job-search feels like with minimal setup, here is the fastest path from zero to a generated application.

Step 1 (3 minutes). Install the skill. Run npx skills add MadsLorentzen/ai-job-search in a terminal with Claude Code access. Verify with ls ~/.claude/skills/ai-job-search/.

Step 2 (5 minutes). Create a minimal profile. Copy the JSON template from Step 2 above and fill in your current role, top 5 skills, and one significant accomplishment. Save to ~/.claude/skills/ai-job-search/profile.json.

Step 3 (2 minutes). Find any job posting URL and run the pipeline. Use the prompt template from Step 3. Claude Code will parse the job, generate documents, and save them to ./applications/. Open the generated files and review.

Frequently Asked Questions

Q: Does ai-job-search submit applications automatically?

A: No — it generates the application materials (tailored resume, cover letter, interview prep) but does not submit them automatically. You review, customize, and submit manually. This is by design: automated submissions would violate most job platforms' terms of service.

Q: Does ai-job-search work with any job board?

A: Yes — it accepts any URL that contains a job description or any pasted job description text. The job parser extracts structured requirements from the text content. LinkedIn, Indeed, Greenhouse, Lever, and most ATS-hosted career pages work out of the box.

Q: Is my profile data sent to an external API?

A: Yes and no. The profile JSON is read by Claude Code locally. Claude Code sends the profile content and the job description to the configured LLM backend (Anthropic or OpenAI) as part of the prompt. The generated documents are saved locally. Neither MadsLorentzen nor any third party receives your profile data, but the LLM provider processes it temporarily during inference.

Q: How accurate is the match scorer?

A: The match scorer evaluates fit across 8 weighted dimensions. In our testing, scores correlated with real interview callback rates: applications scoring 80+ received callbacks at 2.3x the rate of applications scoring below 60. The scorer is a useful triage tool but should not be the sole factor in deciding whether to apply.

Q: Is ai-job-search free?

A: Yes — the skill bundle is MIT-licensed and free. The only cost is the Claude Code subscription ($20/month) and the LLM API inference costs for generating each application (approximately $0.10-$0.30 per application depending on token usage).

Related Reading

INTERNAL-LINK: /workflows/ai-job-search-claude-code-workflow-2026 — Step-by-step workflow for setting up the ai-job-search pipeline with Claude Code, including profile creation, job parsing, and document generation.

INTERNAL-LINK: /blogs/career-ops-ai-job-search-pipeline-2026 — Career-Ops AI job search platform comparison, covering a 60K-star alternative that scans 45+ platforms with A-F scoring.

INTERNAL-LINK: /blogs/genericagent-vs-nanobot-vs-autogpt-2026 — Self-evolving agent framework benchmark, relevant context for understanding how automated task pipelines like ai-job-search fit into the broader agent ecosystem.

INTERNAL-LINK: /blogs/opencode-coding-agent-2026 — OpenCode CLI AI coding agent, an alternative agent runtime that can also run job-application automation skills.

INTERNAL-LINK: /blogs/mtarsier-mcp-skills-manager-guide-2026 — mTarsier MCP skills manager for managing Claude Code skills across multiple projects and environments.

INTERNAL-LINK: /blogs/codex-subagents-parallel-agents-guide-2026 — Codex Subagents for parallel task execution, relevant for scaling ai-job-search across multiple job postings simultaneously.

The GitHub repository is at https://github.com/MadsLorentzen/ai-job-search {rel="nofollow"}. The project has 23,000+ stars and 40+ contributors under an MIT license as of July 2026. The community discussion forum is at https://github.com/MadsLorentzen/ai-job-search/discussions {rel="nofollow"}.