Kill the AI slop in your writing
Free. No account. Pick your surface.
You're using it right now. Paste text, get results. Nothing leaves your browser.
LiveRight-click any text on any page. Scan in the popup or toggle inline highlights directly on the page.
AvailableScan files from the terminal. Pipe text in. Use in CI/CD pipelines. Exit code 1 on findings.
npx sloppy-cli "your text"
sloppy -f article.md
cat draft.md | sloppy
Plug Sloppy into Claude Code, Cursor, Windsurf, or any MCP-compatible AI tool. Your AI gets scan_text, quick_check, and list_patterns as native tools.
// Add to your MCP config:
"sloppy": {
"command": "npx",
"args": ["@sloppy/mcp"]
}
Type /sloppy in any channel to scan text. Right-click a message to check it. Free workspace app.
Zero install. Drag a link to your bookmarks bar. Select text on any page, click the bookmarklet, get results.
AvailableAdd this page to your home screen. Works as a full-screen app on iOS and Android via PWA.
AvailableBuild on top of the engine. Import the core scanner into your own tools, plugins, or workflows.
npm install @sloppy/core
import { scan } from '@sloppy/core'
Tier 1 catches banned words (delve, leverage, utilize). Tier 2 catches signal phrases ("Let's dive in," "Here's the thing"). Tier 3 catches structural patterns (no contractions, sentence stacking, dramatic one-liners).
"Delve" is worse than "enhance." Each pattern has a weight from 1-3. Your total weighted score determines the verdict: Clean, Light Edit, Moderate, or Heavy.
Every hit comes with a specific fix. Not "make it better" — actual replacements and rewrites. You decide what to change.
AI detectors answer "was this written by AI?" — a yes/no guess that's increasingly unreliable. Sloppy answers "where does this sound like AI, and how do you fix it?" Different question. Actually useful.