TermsWatch started from a simple problem: almost nobody actually reads Terms of Service, privacy policies, or long legal documents before agreeing to them. Even when people try, the language is usually too long, complex, and difficult to understand quickly.

We wanted to build something that could explain legal documents in a way that is fast, private, and verifiable.

TermsWatch became a local-first legal document assistant that works across websites, PDFs, scanned documents, and photos of physical paperwork. It gives short plain-English explanations, identifies important clauses, and—most importantly—links every explanation back to the exact source text so users can verify what the system is saying.

We also added policy comparison, allowing TermsWatch to show what changed between old and new versions of online terms.

One of the biggest challenges was making OCR and source highlighting work reliably across desktop and Android while keeping processing local. We also built safeguards so unsupported AI claims are rejected rather than shown as facts.

The final result is a privacy-focused tool designed around one idea:

Know before you agree.

Built With

  • android/java
  • capacitor
  • chrome-manifest-v3
  • eslint
  • indexeddb
  • onnx-runtime-web
  • pdf.js
  • react
  • smollm2-360m-instruct
  • tailwind-css
  • tesseract.js
  • transformers.js
  • typescript
  • vite
  • vitest
  • web-workers
Share this project:

Updates

Submission history