The only AI content detector that runs 100% on your infrastructure. Zero external API calls. Zero data leaves your network.
Visit manuscript.dev → | View on GitHub →
Why I Built This#
Every AI detection tool on the market sends your content to someone else’s servers. For healthcare, legal, finance, and government organizations, that’s a non-starter. Manuscript runs entirely on your own infrastructure.
What It Detects#
| Content Type | Accuracy | Method |
|---|---|---|
| Text | 90% | Linguistic analysis — sentence variance, vocabulary richness, contraction patterns |
| Images | 50% | Forensic signals — EXIF metadata, sensor noise, compression artifacts |
| Audio | 46% | Spectral analysis — FFT, MFCC computation, AI signature detection |
| Video | In Progress | Temporal patterns, encoding signatures, deepfake detection |
Key Features#
Privacy-First — No data ever leaves your network. HIPAA, SOC2, PCI compliant by design.
- Self-Hosted — Docker, Kubernetes, or bare metal. Your infrastructure, your rules
- Blazing Fast — Sub-10ms response times. Pure Go, zero external dependencies
- Multi-Modal — Text, images, audio, and video detection in a single API
- Open Source — MIT licensed. Audit the code yourself
Quick Start#
docker run -d \
--name manuscript \
-p 8080:8080 \
--restart unless-stopped \
manuscript/manuscript:latestThen verify content:
curl -X POST http://localhost:8080/verify \
-H "Content-Type: application/json" \
-d '{"text": "Your content to analyze..."}'Built For#
- Healthcare — HIPAA compliance requires data stays on-premise
- Legal — Attorney-client privilege demands zero external calls
- Finance — SOC2/PCI requirements for content verification
- Government — Air-gapped network compatible
- Education — No per-seat licensing, unlimited scale
Tech Stack#
