Cortex FAQs

Question 1

What is Cortex and what does it do?

Accepted Answer

Cortex is a local-first memory OS for AI coding agents, functioning as a Model Context Protocol (MCP) server. It enhances Claude Code by layering intelligent memory features like hybrid search, a knowledge graph, lifecycle management, and per-turn context briefs over your existing Engram database, without altering core data.

Question 2

Is Cortex a cloud-based solution?

Accepted Answer

No, Cortex is explicitly designed as a local-first tool. It operates entirely on your machine, leveraging SQLite, `sqlite-vec`, and Ollama. This architecture ensures zero cloud dependencies, zero cost, and complete data privacy, running solely on your local environment.

Question 3

How does Cortex manage context for AI agents?

Accepted Answer

Cortex employs a sophisticated three-layer pipeline for context management: FTS5 full-text search, vector similarity search, and knowledge graph expansion to surface related observations. These results are then enriched by lifecycle management, which filters deprecated/superseded entries and applies a composite score based on search relevance, confidence, and recency, ultimately delivering highly relevant context through `cortex_preflight` briefs.

Question 4

What makes Cortex unique compared to other memory tools for AI?

Accepted Answer

Cortex stands out with its powerful hybrid FTS5 + vector search employing Reciprocal Rank Fusion, a comprehensive knowledge graph for entity extraction and graph expansion, and advanced lifecycle management with confidence scoring and staleness detection. It's also entirely local-first (using SQLite and Ollama) and MCP native, requiring zero cloud dependencies or additional infrastructure.

Question 5

How does Cortex improve AI coding agent performance?

Accepted Answer

Cortex significantly boosts agent performance through smarter retrieval via hybrid FTS5 + vector search, entity awareness from its knowledge graph, staleness detection, and automatic context injection. The `cortex_preflight` feature provides concise, token-budgeted context briefs at the start of each turn, ensuring agents have the most relevant information.

Cortex

Cortex

주요 기능

사용 사례

주요 기능

사용 사례