ai-job-search-opencode

The job search that runs on your machine.

Fork it. Profile it. Let AI handle the grind. Evaluate postings, tailor your CV, write cover letters, prep interviews — all from your terminal.

Community fork of MadsLorentzen/ai-job-search (22.9k stars), adapted from Claude Code to OpenCode. All credit for the original framework goes to Mads Lorentzen.

Does it actually work?

The original framework was built by a geophysicist who got laid off in late 2025. He ran this exact workflow weekly on his own career:

69 tailored applications → 20 first interviews → 1 signed contract.

He started as an AI engineer in June 2026. The framework got him hired. Now it's yours.

Read the full story on LinkedIn · Buy him a coffee

How it works

Three agents, one pipeline:

What makes this different

Drafter-reviewer separation. Two agents, two contexts. The drafter writes; a fresh reviewer agent researches the company and critiques every draft. No single-pass generic output.

PDF verification loop. LaTeX looks fine in .tex and broken in the PDF. This workflow compiles with lualatex + xelatex, inspects every page, and fixes orphaned titles, font mismatches, and page overflows before you ever see the file.

ATS-aware keyword verification. Recruiters don't read your PDF — parsers do. pdftotext extracts the text layer and verifies contact details, reading order, and keyword coverage against the posting. Gaps stay visible, never stuffed.