Manuscript — AI content detection that runs on your hardware.
Pure Go. 4-modal detection (text, code, image, audio). On-prem. Your data never leaves your network. MIT-licensed and audit-able.
Four modalities. One binary. Your hardware.
Manuscript catches AI-generated text, code, images, and audio without sending a byte to a third-party API. The detection logic is open source, so your compliance team can read every check before approving it.
Clinical notes, legal filings, regulatory submissions, financial-services documentation.
Compliance reviews where attribution and provenance matter. Spot AI-generated commits.
Synthetic medical imaging, document forgeries, KYC imagery checks. Pixel-level analysis.
Voice-cloned customer-service recordings, deepfake screening for regulated workflows.
By the numbers
What Manuscript looks like in production.
4
Modalities — text, code, images, audio
0
Bytes sent to third-party APIs — runs on your hardware
100
% MIT-licensed and audit-able
Why on-prem matters
HIPAA-aligned deployment. No third-party API calls. No "we anonymize before sending." The binary runs where you run it.
Read the detection logic. Run your own benchmarks. Submit a PR. Manuscript can pass a regulator's scrutiny because the code is in the open.
Pure Go. No Python runtime, no GPU required for the baseline detector. Air-gapped environments supported.
Mindtrades engineers ship Manuscript on real client engagements. The detector you run is the one we use.
Who runs Manuscript
healthcare system, NDA
law firm, NDA
FinTech compliance team, NDA
MIT-licensed and free, or have us deploy and operate it.
Self-serve: Docker pull or compile from source — no signup, no telemetry. For regulated industries that need a managed on-prem deployment with named accountability and signed audit reports, book a 30-min call below.
