ResuRank is an open-source desktop application for evaluating resume-to-job fit without sending either document to a third party. Upload your resume PDF once and ResuRank stores it locally. Paste any job description and get a 0–100% match score within seconds, computed by two independent methods that complement each other: semantic similarity (vector embeddings) and keyword similarity (TF-IDF with rarity weighting), blended with a divergence adjustment that protects against false positives.
Everything runs locally. The ~25 MB embedding model is downloaded once on first launch and cached in your user data directory. Your resume, settings, term boosts, and job descriptions never leave your machine — no cloud, no API keys, no usage limits, no telemetry.
It also exposes its scoring logic through an MCP (Model Context Protocol) service, so AI assistants can run resume-to-job evaluations directly through tools like Claude Code.
Features
- Hybrid scoring — semantic embedding (70%) + keyword TF-IDF (30%) with divergence adjustment
- Local inference — runs entirely on your device, no internet required after first run
- Term boosting — weight specific keywords to reflect skills you want to emphasize
- Critical missing keywords — flag must-have terms; their absence reduces the score with adjustable importance tiers
- Stopword exclusion — customize the word list ignored during scoring
- Score breakdown — see embedding, TF-IDF, overlap bonus, and divergence penalty separately
- Language detection — warns when a job description appears to be in a different language
- PDF resume parsing — upload once, reuse for every job
- Auto-update — signed updates delivered via electron-updater on macOS and Windows
- Score tiers — Poor fit / Fair / Good / Great fit at a glance
- MCP service — connect ResuRank to AI assistants via Model Context Protocol