Cortex FAQs

Question 1

How do I get started with Cortex and what platforms does it support?

Accepted Answer

You can install Cortex via Homebrew (macOS), npx (any platform), or a curl script. After installation, run `cortex init` in your project directory. Cortex supports various platforms and integrates with Claude Code via the Model Context Protocol.

Question 2

What is Cortex and how does it solve the problem of Claude Code forgetting context?

Accepted Answer

Cortex is a local MCP server that provides persistent memory for Claude Code. It automatically captures and injects session context, project architecture, decisions, and preferences, ensuring your AI assistant remembers everything across sessions and eliminates the need for constant re-explanation.

Question 3

How does Cortex ensure only high-quality and relevant memories are stored?

Accepted Answer

Cortex employs a robust Quality Gate with 7 validation rules to block duplicates, sensitive data, and low-quality saves. Additionally, an AI Session Summarizer reviews completed sessions to automatically capture any important context Claude Code might have missed.

Question 4

Is my code or data sent to external servers when using Cortex?

Accepted Answer

No. Cortex is 'local first.' It runs entirely on your machine, and your memories are stored in a local SQLite database. Data only leaves your machine if you explicitly enable the optional Turso-powered multi-machine sync, which you fully control.

Question 5

What kind of information does Cortex store, and how is it organized?

Accepted Answer

Cortex organizes memories into 6 distinct types: Decision, Context, Preference, Thread, Error, and Learning, each designed for specific information lifecycles. It uses a 4-layer strategy for project detection to associate memories accurately with your work.

Cortex

Cortex

主な機能

ユースケース

主な機能

ユースケース