About this project

slop-chop is a tool that strips common AI writing patterns from text, such as em-dashes, overused words like 'comprehensive' and 'leverage', and stock phrases like 'In summary'. It performs a deterministic, local rules pass that swaps characters, drops flagged words, rewrites stock phrases, applies custom regex patterns, enforces spelling dialects, and tidies punctuation. It understands Markdown, leaving code blocks and inline code untouched. An optional rewrite pass uses a language model (Anthropic or any OpenAI-compatible API, including local Ollama) for deeper changes like reworking sentences to avoid semicolons or adjusting voice. The tool offers multiple interfaces: a CLI (`slop-chop fix`, `check`, `score`, `attack`, `voice`), a GitHub Action for CI, an Obsidian plugin, an MCP server, a VS Code/JetBrains extension via LSP, a web app at slop-chop.com, a Node package (WASM), a Go library, an HTTP API, and a Slack integration. It supports custom profiles (JSON) and presets for different contexts (corporate, academic, marketing, etc.). The `score` command rates text density of tells from 0 to 100, and `check` can be used as a CI gate. The `voice` feature helps maintain personal writing style by protecting preferred words and detecting drift. The engine is benchmarked against a labeled corpus to ensure detection quality.