code cleanliness scores for vibe-coded apps
paste a github repo. get a cleanliness score, a breakdown of exactly what's messy and where, and ready-to-paste claude prompts that tidy it all up — in under a minute. free · no card · nothing written to disk · powered by claude.
we stream the archive in memory — no clone, no install, nothing touches disk. pro plans scan private repos via github.
one number, six weighted subscores, and every finding ranked by severity with file + line — plus a claude-written summary of what a tidy-up would change.
each finding ships a claude-ready prompt with behaviour-preserving constraints. paste it into claude code, review the diff, re-scan.
1,500-line files, flat folder trees, and a root directory where config, source and one-off scripts all blur together.
six-deep nesting, lines that run off the screen, logic nobody can hold in their head — the stuff reviews silently skip.
the same block pasted into four files with one variable renamed. every copy is a bug you will fix three times.
commented-out functions, stale TODOs, console.logs from that one debugging session in march.
tabs and spaces at war, three filename conventions, two package managers — phantom diffs forever.
node_modules committed, .DS_Store everywhere, no README, no lockfile, no .gitignore, zero tests.
one free scan · analysis by claude haiku · full report, every finding, every claude prompt · shareable report + live readme badge.
claude opus 4.8 reads your whole repo, every finding double-checked to filter false positives · private repos · one-click fix PRs + github issues export · weekly monitoring + alerts · 300 credits/mo · also $9/wk or $228/yr. failed scans are refunded automatically.
one number, 0–100, built from six weighted subscores (structure, readability, duplication, dead code, consistency, hygiene). every finding deducts points by severity, so the score moves when you actually clean things up — re-scan and watch it climb.
no. the repo archive is streamed in memory, never written to disk, and discarded the moment the scan ends. the only thing stored is your report. analysis runs on the claude api, which does not train on your data.
every finding ships a ready-to-paste prompt for claude code (or any coding agent): the issue, the file and line, why it matters, what to change, and behaviour-preserving constraints. there is also one tidy-up-everything plan covering the whole report.
yes, on pro — connect github and cleanvibes fetches your private repos through your own token. free plans scan public repos only.
it's tuned for vibe-coded apps — fast ai-built codebases that never got a tidy-up pass — but the checks are just as true of any rushed human codebase.