Marcut

Professional, Local-First Document Redaction.

Marcut is a native macOS application for legal and professional document redaction. It combines a deterministic rules engine with optional local AI (Ollama) to identify and redact sensitive information (PII) from Microsoft Word (.docx) documents, producing "Track Changes" redlines plus JSON audit and scrub reports.

Platform: Marcut is a macOS-only application (Apple Silicon and Intel, Universal2). The marcut Python CLI can run on other platforms with a local Ollama install, but is developed and tested on macOS.

See Technical Architecture for a system design overview.

🚀 Key Features

• Local-First & Private: All processing happens on your device. No cloud uploads.
• Self-Contained Runtime: Embedded Python (BeeWare) and local AI service; no system installs required.
• Dual-Engine Power:
  • Rules Engine: Instant, regex-based detection for structured data (SSN, Phone, Email, etc.).
  • AI Engine: Context-aware entity recognition (Names, Organizations) using local Ollama models.
• Professional Output: Generates standard DOCX review artifacts with redacting edits marked as "Track Changes"; use Send Final Redacted Copy to accept Marcut redaction changes and scrub metadata in a separate share copy, or intentionally send a review copy after confirming Track Changes and metadata may remain recoverable.
• Audit Ready: Produces JSON redaction reports and metadata scrub reports for verification; reports may include raw detected text and document metadata.
• Fail-Safe by Design: Pre-flight checks (writable destination, free disk space), clear plain-English errors, live per-document progress, and fail-closed handling if the AI can't fully analyze a document (no silently incomplete redactions).
• App Store Ready: Fully sandboxed and code-signed architecture.

📸 Screenshots

Live AI progress	Redaction settings

Redaction streams progress per chunk as the AI model works, with a live heartbeat and time estimate — and finishes with a clear per-document status:

📖 Documentation

• User Guide: Installation, usage instructions, and troubleshooting.
• Developer Guide: Deep dive into the Swift+PythonKit architecture, build system, and contribution workflow.
• Technical Architecture: High-level system design and component interaction.
• Security Policy: Vulnerability reporting and security model.

🛠️ Quick Start (Development)

To build Marcut from source (macOS required):

1. Clone the repository:

   git clone https://github.com/legalmarc/marcut.git
   cd marcut

2. Configure: Copy the example config and edit if needed (mostly for signing identities).

   cp build-scripts/config.example.json build-scripts/config.json

3. Build: Use the TUI orchestrator (single entrypoint) to create a local dev build.

   ./build_tui.py

   Then choose Build Workflows → Dev Fast + Test Prep.

4. Run: Launch the built app:

   open src/swift/MarcutApp/.build/arm64-apple-macosx/debug/MarcutApp

For detailed build instructions, see the Developer Guide.

📦 Distribution

Marcut ships as a Developer ID signed and notarized DMG via GitHub Releases. Homebrew cask and PyPI distribution are not yet available.

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines on code style, testing, and the pull request process, and our Code of Conduct.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details. Bundled third-party components remain under their own licenses - see THIRD_PARTY_LICENSES.md.

---

Built with SwiftUI, PythonKit, BeeWare, and local Ollama.