TokenHush Lite — Local LLM privacy gateway
Auto-mask keys, private keys, emails and credit cards before they reach any model
- Problem
- A developer lost over ¥10,000 when a cloud key leaked during vibe coding, and others worry about sending sensitive data into models. An open-source local LLM gateway already replaces secrets with placeholders and restores them on the way back, proving the pain is real—but it still requires setup for non-technical users.
- Solution
- A ready-to-run local privacy gateway: proxy OpenAI/Claude and other endpoints, detect and replace API keys, JWTs, PEM private keys, emails and credit cards, then restore them in the response. It stays invisible to agents and coding tools, with one-click launch, visual rule management and request logs.
- Users
- Indie developers and small teams using Cursor, Claude Code, Codex and similar AI coding tools, plus freelancers handling client data
- MVP
- One-click local proxy; built-in regex rules for common secrets; placeholder replace/restore; support major LLM upstreams; simple log and rule toggle UI
- Revenue
- Open-source core plus paid desktop GUI and team policy sync (one-time or subscription); enterprise tier with audit export and self-hosting
- Why now
- AI coding tools have massively expanded secret exposure while cloud gateways don't serve individual developers—a natural opening for a lightweight local tool