What is Sci?

Sci is a sovereign cognitive identity layer for the AI tools you already use. It’s the first product from DancingBits, and the alpha targets one tool deeply: Claude Code.

It does two things that normally pull in opposite directions — it makes the model know less about you, and more about your work.

The problem it solves

Coding assistants are remarkable, but using one means narrating your life to a remote model: your name in a commit, a client’s domain in a stack trace, an API token you pasted by accident. That text leaves your machine, travels networks you don’t control, and may be retained or used to train. Meanwhile the useful context — the decisions you made, the conventions of your project, the things you explained yesterday — evaporates the moment the session closes.

Sci splits those two flows apart. The identifying details get stripped out before they leave. The working context gets captured and carried forward.

1. Privacy proxy

Every request from Claude Code passes through a local proxy before it reaches Anthropic. Sci detects your real identifiers — name, email, and other PII — and replaces them with stable placeholder tokens ([PERSON_1], [EMAIL_1]). “Stable” matters: the same person maps to the same token every time, so the model can still reason about “the user” coherently — it just never learns the user is you. When the reply returns, Sci swaps the tokens back. You read a normal conversation; Anthropic only ever saw the tokens.

2. Persistent memory

Sci runs a local memory store and exposes it to Claude Code through four MCP tools. Claude can recall facts about you, your projects, and past decisions, and store new ones as you work. Open a fresh session in a different repo tomorrow and the relevant context is already available — you stop re-explaining who you are and what you’re building.

Sovereignty by default

The principle underneath both features is simple: your data stays yours.

• Memory lives in a local SQLite database on your machine.
• Embeddings are computed on-device — no text is shipped to a hosted embedding API to be indexed.
• Sci never holds your credentials. The proxy forwards whatever login Claude Code already uses — your OAuth session (Claude Pro or Max) or your own API key — without storing or inspecting it.
• There is no Sci account and no Sci server. Nothing about you is phoned home, because there’s nowhere for it to go.

What the alpha proves

v0.1.0 is about establishing the core pipeline end-to-end on real hardware: anonymize → forward → deanonymize, plus memory injection, working transparently inside Claude Code. Auto-update, a GUI, multi-provider routing, and cloud sync come later.

Ready? Head to the install guide, or look at how it works under the hood.