Purpose
WSC (Writing Style Checker) is a prose style checking and AI-generated trace detection tool. It is used to find words, phrases, and sentence structures that appear excessively in AI-generated text, and each flag is accompanied by published corpus research evidence. It can also catch classic writing problems, including vague wording, passive voice, repeated words, overly long sentences, nominalizations, vague expressions, and filler adverbs.
Main Capabilities
WSC provides 8 detectors, covering the above writing problems and AI traces. The detection word lists are sourced from Matt Might's shell script, with additional entries added; the AI trace portion is based on published corpus research, including Kobak et al. 2025 (Science Advances), Juzek & Ward 2025 (COLING), Liang et al. 2024 (Stanford), and Reinhart et al. 2025 (PNAS), supplemented by a catalog of AI writing signs maintained by Wikipedia editors and reports from AI detection vendors.
Detectors can be customized via .wscrc.json; all fields are optional, missing fields use default values, and a JSON Schema is provided for autocompletion in VS Code.
How to Use
WSC is available in multiple forms:
- Web editor: real-time highlighting, with inline fix buttons for all 8 detectors. The web editor runs in the browser and does not send text to any server. The online address is wsc.theserverless.dev.
- HTTP API: POST text, with an optional configuration object, and returns a structured JSON response. The core endpoint is POST /api/check, used to analyze text; there are also endpoints that return API documentation and the detector list, as well as a health check endpoint. Limits: up to 100,000 characters per request, CORS enabled for all origins. The API and MCP endpoints only process the text you explicitly send to them.
- MCP server (remote): connect AI assistants via Streamable HTTP transport, with no installation required.
- MCP server (local): use a stdio-based server via wsc-mcp on npm, including check_file for analyzing files on disk and automatically discovering .wscrc.json.
- CLI: check files from the command line via wsc-lint.
- GitHub Action: run checks in CI, using ::warning annotations.
Local development: after running, visit http://localhost:5173，API is located at /api/check, MCP is at /mcp, and the health check is at /health. Deployed as wsc.theserverless.dev on Cloudflare Worker.
Use Cases
Suitable for scenarios that require checking text style, improving writing clarity, or identifying AI-generated traces in text. It can be used for real-time web editing, API integration, AI assistant writing checks, local file analysis, and CI workflows.
Limitations and Notes
WSC flags patterns that research shows appear excessively in AI-generated text and cites sources; it cannot and does not prove authorship. Classifier-based detectors have documented risks of misjudgment: a Stanford study found that seven classifiers incorrectly flagged 61% of articles written by non-native English speakers. WSC avoids this trap by design—each flag is a specific, explainable revision suggestion that can improve the text regardless of who or what generated it.
