TAU-HOME.COM
LOADING

ai-job-search: A Local Career Automation Framework Built on Claude Code

Explore ai-job-search, an open-source framework leveraging Claude Code on your local machine to scrape job listings, evaluate fit, generate tailored LaTeX resum

tau · November 15, 2026

#claude-code #ai-job-search #career #ai-agents #latex

ai-job-search: A Local Career Automation Framework Built on Claude Code

Job hunting requires navigating numerous repetitive workflows, from scouting postings across multiple boards to customizing resumes and preparing for multi-round interviews. Emerging from the open-source community, ai-job-search transforms Anthropic's terminal agent, Claude Code, into a full-stack career assistant operating entirely on the user's local machine. Originally created in March 2026, the repository has accumulated over 71,500 stars and 23,100 forks, publishing its v1.7.5 release on November 9, 2026 under a standardized weekly release cycle.

ai-job-search GitHub workflow diagram and Claude Code job search pipeline

Image source: https://github.com/ai-job-search-dev/ai-job-search

Conventional job platforms and cloud-hosted generative AI services often require uploading sensitive career records, compensation figures, and personal identifiers to remote third-party servers. Maintained under the ai-job-search-dev organization, this MIT-licensed framework keeps profile data and draft documents in local storage, orchestrating an end-to-end job application funnel through structured terminal commands.

A Six-Stage Command Lifecycle from Setup to Tracking

The operational core of ai-job-search consists of six slash commands executed inside the Claude Code Command Line Interface (CLI). Instead of requiring users to construct ad-hoc prompts, the framework guides autonomous sub-agents through structured routines:

  • /setup (Profile Construction): Ingests user career background into structured data files. It supports scanning a local documents/ directory containing resume PDFs, reference letters, diplomas, and past applications. It also allows pasting raw resume text or conducting an interactive agent interview to establish skill sets and experience levels.
  • /scrape (Job Discovery): Queries connected recruitment portals for listings matching the profile, deduplicates entries, and sorts opportunities by relevance. Scraped history is tracked locally, dropping previously reviewed listings to avoid bloating Claude Code's active context window.
  • /rank (Batch Fit Scoring): When a search returns dozens of matches, this command evaluates role criteria against user qualifications in bulk, surfacing an objective, prioritized shortlist.
  • /apply (Document Drafting and Compilation): Accepts a job URL or raw description text. It evaluates candidate fit, drafts tailored resumes and cover letters, submits drafts to a secondary reviewer agent, and compiles clean LaTeX output into ready-to-send PDF files.
  • /interview (Interview Preparation): Analyzes target company requirements, anticipates role-specific technical and behavioral questions, and conducts interactive mock interviews with constructive feedback.
  • /outcome (Funnel Analytics): Records screening decisions, interview milestones, and final offers to monitor career search progress over time.

While the bundled portal search skills cater primarily to the Danish market (such as Jobindex and Jobnet), the core drafting and evaluation pipeline operates completely language- and country-agnostic. Users targeting other jurisdictions can integrate regional job sites using the /add-portal command.

Runtime Requirements and the Dual-Agent Review Pipeline

Running ai-job-search requires four foundational tools configured on the host machine. First, Claude Code CLI must be authenticated; because Anthropic offers no free tier for Claude Code, users need an active paid plan (Claude Pro, Max, or Team) or Anthropic API credits. Second, Python 3.10+ is required to execute repository tooling. Third, the Bun runtime is necessary to manage CLI portal adapters under .agents/skills/*/cli/. Fourth, a complete LaTeX distribution (such as TeX Live, MacTeX, TinyTeX, or MiKTeX) must be available.

The document generation engine under /apply is built around dual TeX engines to avoid typesetting anomalies. Resumes are compiled with lualatex, avoiding font-expansion issues common in modern TeX distributions, while cover letters compile through xelatex using the fontspec package and custom styling classes. An optional pypdf installation allows the framework to verify machine readability against automated Applicant Tracking Systems (ATS), falling back to Poppler's pdftotext or visual keyword checks when unavailable.

Quality assurance is enforced through a two-stage drafter-reviewer architecture. The initial drafting agent generates tailored descriptions based strictly on verified profile facts. A secondary reviewer agent then cross-examines the draft against the original job requirements and candidate records, preventing artificial hallucinations or unverified credential claims before LaTeX compilation begins.

Critical Setup Rules: Private Repository Cloning and Input Integrity

Managing the repository requires strict attention to git configuration and data privacy. Under GitHub's platform rules, forking a public repository results in a permanently public fork. Because running /setup writes sensitive personal details—including legal names, home addresses, phone numbers, previous employers, and salary expectations—into tracked files, pushing changes to a standard public fork risks immediate public exposure of personal data.

To protect personal information, job seekers must initialize a private repository copy rather than creating a public fork:

# Clone the official repository locally
gh repo clone ai-job-search-dev/ai-job-search
cd ai-job-search

# Rename origin remote to upstream for tracking future updates
git remote rename origin upstream

# Create a private GitHub repository and push your local copy
gh repo create ai-job-search --private --source=. --remote=origin --push
gh repo set-default <your-github-username>/ai-job-search

# Install portal CLI dependencies
python3 tools/install_portal_clis.py

This private mirror structure ensures that confidential files remain protected within a private repository, while upstream bug fixes and weekly releases from ai-job-search-dev/ai-job-search can be pulled seamlessly.

Users must also recognize security boundaries when handling raw job descriptions. The framework applies deterministic integrity checks to filter invisible unicode characters, obfuscated hyperlinks, and prompt injection patterns embedded within postings. However, because LLM defenses operate at the instruction level rather than within an isolated operating system sandbox, users must review all generated PDF documents and verify posting details before sending applications to prospective employers.

Sources

Impact-Site-Verification: 9b3ebd10-6c45-44cf-a014-5b8c19b7b285