Quick Start

Get Sci running and connected to Claude Code in under 5 minutes.

Prerequisites

• Docker Desktop
• Node.js 18+
• Claude Code (or any MCP-compatible agent)

Install

git clone https://github.com/sovereign-cognitive-identity/sci
cd sci
npm install

Start the database

docker compose up -d

This starts Postgres 16 with pgvector. The schema is applied automatically on first run.

Build

npm run build

First-run setup

sci setup

This runs a 4-step wizard:

1. Verifies the database connection
2. Checks the schema
3. Shows current memory counts
4. Generates a trusted token for Claude Code

Copy the claude mcp add command it outputs and run it.

Restart Claude Code

After running claude mcp add, restart Claude Code. The sci MCP server will be available with four tools:

• memory_recall — semantic search across all memory
• memory_store — store a new episodic memory
• memory_identity — retrieve your identity facts
• memory_status — health check

Seed your memory (optional but recommended)

Export your Claude conversation history from claude.ai/settings/account → Privacy → Export Data.

Then:

sci import --claude ~/Downloads/conversations.json

This seeds episodic memories and runs identity extraction, so Sci immediately knows your projects, preferences, and context.

Verify the privacy guarantee

Before trusting Sci with real conversations, run:

node demo/privacy-demo.mjs

This proves in under 10 seconds that your real name never appears in outbound text. All 6 checks must pass.

What's next

• Connect additional agents — scope Cursor to your work profile, give a third-party agent read-only access
• Choose a storage backend — move your data to Dropbox, S3, or iCloud
• Set up nightly consolidation — let Sci learn from your conversations automatically