Sonomos
AI & ML interests
None defined yet.
SONOMOS Privacy at the point of creation.
We build local-first privacy infrastructure for AI. Our tools detect and mask sensitive databefore it reaches any AI model. Everything runs on your device. Nothing leaves your machine unscreened.
The problem
Every prompt is a potential data leak. Developers paste code with API keys. Users share documents with client names. Teams discuss strategy with internal numbers. AI sees everything, and there is no undo.
Our approach
We intercept at every surface where data meets AI — browsers, desktops, MCP tool calls, agent workflows — and apply detection and masking locally, before transmission. No cloud. No server-side scanning. No trust required.
Public Projects
Canary — PII exposure counter for Claude Code
Counts every piece of sensitive data you expose across all sessions. 16 regex detectors with real checksum validation plus 70+ semantic categories via Claude self-scan. Local-only, zero config. The number only goes up.
/plugin marketplace add sonomos-ai/Canary-Plugin
/plugin install canary@sonomos
What We're Building
| Surface | What it does | Status |
|---|---|---|
| Locke | On-device sensitive data detection with format-preserving encryption, image pipeline (face/NSFW/OCR redaction), cognitive firewall | In development |
| "Named Entity Recognition Validation Engine" | An open-source, deterministic NER benchmark testing harness for evaluating sensitive data detection | In development |
| Canary | PII exposure counter for Claude Code — shows what you have already leaked | Public |
Everything local. Everything on-device. No exceptions.
Contributions welcome.