Home

MemoryWeft Wiki

Local‑first graph memory for MCP — where events are the edges.

MemoryWeft (MWeft) is a long‑term memory for AI that runs on your machine. This wiki covers the ideas and the data model. For installation and tool usage, see the repository docs linked below.

English

• Concepts — what MWeft is, why events are edges, the two graphs, embeddings, tags & domains.
• Data model — entities, events, tags, relationships, and how membership is derived.

한국어

• 개념 (Concepts) — MWeft가 무엇인지, 왜 이벤트가 엣지인지, 두 그래프, 임베딩, 태그·도메인.
• 데이터 모델 (Data model) — 엔티티·이벤트·태그·관계, 소속이 derive되는 방식.

Repository docs

• Install — installation, backends (SQLite / PostgreSQL), embedding options.
• Prompt guide — how an AI searches and saves through the MCP tools.
• README · README (한국어)