Noema

•
25
•

Provides AI agents with a persistent, structured memory layer to record observations, decisions, and intentions.

Noema acts as an intentional memory layer for AI agents and the humans collaborating with them, offering a durable and organized space to store knowledge, choices, observations, and plans. It operates locally, saving every memory as a plain markdown file with an index in a SQLite database, ensuring full user control without relying on cloud services or API keys. The tool supports various 'Trace' types to categorize memories and includes powerful CLI and MCP server capabilities for management and integration.

주요 기능

012 GitHub stars
02Local-first, persistent memory for AI agents using plain markdown files and a SQLite index.
03Supports diverse 'Trace' types like facts, decisions, preferences, skills, and intentions.
04Offers a comprehensive Command Line Interface (CLI) for managing memories (add, list, search, edit, archive, sync).
05Functions as an MCP (Model Context Protocol) server, providing AI tools direct access via stdio or Streamable HTTP.
06Enables federation for peer-to-peer memory synchronization with shared-key authentication.

사용 사례

01Equipping AI agents with durable, structured memory to enhance their long-term knowledge retention and context awareness.
02Facilitating human-AI collaboration by providing a shared, auditable record of observations and decisions.
03Integrating with MCP-compatible AI clients and development environments, such as Claude Code, Zed, and GitHub Copilot.