Remnic

•
235
•

Provides open-source, local-first memory and context for AI agents, ensuring consistent recall across various tools and sessions.

Remnic is an open-source, local-first memory and context solution for AI agents, designed to ensure agents consistently understand users, projects, and context across sessions. It stores every memory as plain markdown files with YAML frontmatter on your local disk, eliminating cloud dependencies. By combining a robust memory backend with a human-readable context substrate, Remnic offers sharp hybrid search retrieval, automatic knowledge extraction, and the ability to compound knowledge over time, ensuring agents never forget important details or repeat mistakes.

主要功能

01Unified memory store accessible by various AI tools and agents including OpenClaw, Claude Code, and any MCP client.
02Advanced hybrid retrieval combining BM25, vector search, reranking, graph recall, and memory-worth scoring.
03165 GitHub stars
04Local-first storage using plain markdown files with YAML frontmatter for human readability and version control.
05MIT licensed, allowing free use, modification, and distribution.
06Automatic knowledge extraction from conversations and intelligent context injection when needed.

使用案例

01Providing consistent identity, project, and preference context to AI agents across different sessions and tools.
02Recalling past debugging findings, research, and decisions to avoid repetitive work and accelerate project completion.
03Enabling human-governed correction loops for agent beliefs, ensuring accurate and up-to-date knowledge propagation.