ClaudeMark v2.0.0 — Multi-AI Watermark & Provenance Forensics
ClaudeMark v1.0.0
ClaudeMark is a local-first, open-source toolkit for AI watermark research, provenance inspection, Unicode forensics, and document cleaning.
Highlights
- Claude-focused statistical research detector
- Kirchenbauer red/green list detector
- SynthID-style experimental research adapter
- Generic entropy and burstiness analysis
- Unicode and invisible-character forensics
- C2PA provenance inspection
- EXIF, XMP and IPTC analysis
- AI-generator footprint detection
- PDF, DOCX, ODT, HTML, Markdown and TXT support
- PNG, JPEG, WebP, SVG, AVIF and HEIC support
- Batch inspection and cleaning
- Forensic text diff
- Safe text normalization
- JSON, Markdown and terminal reports
- Local REST API
- Interactive web dashboard
- Docker support
- Benchmark and calibration framework
Scientific positioning
ClaudeMark is a research and forensic analysis toolkit.
Detector results are statistical signals and should not be interpreted as definitive proof of AI authorship or model origin. The SynthID-style component is experimental and is not an official Google SynthID detector.
Privacy
ClaudeMark is designed for local-first execution. Core analysis and cleaning operations do not require external cloud services or telemetry.
Automated tests include zero-egress checks for core operations.
Validation
- 48 automated tests passing
- CLI commands verified
- File inspection and cleaning tested
- Unicode edge cases tested
- Malformed/corrupted file handling tested
- Zero-network-egress tests included
Quick Start
python claudemark.py version
python claudemark.py capabilities
python claudemark.py analyze --text "Example text"
python claudemark.py inspect README.md
python claudemark.py normalize --text "Example text"
**Full Changelog**: https://github.com/karthikrshet/ClaudeMark/commits/v1.0.0