basicmachines-co/basic-memory

What it solves

Basic Memory provides a persistent, two-way memory system for AI agents and humans. It solves the problem of ephemeral LLM conversations by allowing AI assistants to read and write structured knowledge into plain Markdown files on a user's disk, creating a long-term memory that persists across sessions and different AI clients.

How it works

The system uses the Model Context Protocol (MCP) to connect AI clients (like Claude, Cursor, or ChatGPT) to a local or cloud-hosted directory of Markdown files. These files act as "Entities" containing observations (facts) and relations (wiki-links) that form a knowledge graph. The AI can search these notes using semantic vector search (via FastEmbed) or keyword search, and can update the notes in real-time. Because the data is stored in plain text, users can edit the notes manually in editors like Obsidian or VS Code, and the AI will see those changes.

Who it’s for

Developers and power users who want their AI assistants to maintain a running log of project decisions, learnings, and personal knowledge without being locked into a proprietary database or complex RAG infrastructure.

Highlights

• Local-first and Two-way: Knowledge is stored as plain Markdown files that both humans and AI can read and write.
• MCP-native: Compatible with any AI client that supports the Model Context Protocol.
• Semantic Search: Supports hybrid full-text and vector search with optional cross-encoder reranking for higher precision.
• Knowledge Graph: Uses wikilinks to create a traversable graph of information that AI agents can follow.
• Cros-device Sync: Optional cloud hosting for synchronization across mobile, web, and desktop devices.