Memind

Memind is a long term memory system for AI agents and applications. Rather than storing memory as a flat list of isolated facts, it extracts, organizes and evolves knowledge from conversations and raw content into structured memory that an agent can retrieve as concise context. The design keeps the original source material as a raw data layer, then derives structured memory items connected through an item graph of entities, aliases, mentions and links. Related items are consolidated into a tree of leaf, branch and root insights, and grouped into memory threads that follow a topic across many sessions. Memory items carry temporal fields, so queries with time constraints can be answered differently from timeless facts. Memory is split into user and agent scopes, so a coding or task agent can remember tool usage, durable instructions, reusable playbooks and resolved problems separately from facts about a person. The documentation describes extraction from conversations, streaming messages, documents, images, audio, tool calls and custom raw content. Conversation extraction is built into the core module, while documents, images, audio and tool calls require matching plugins. The site also covers integrations, an API reference and release notes, and has a section for the open source project. The documentation publishes no pricing page or plan prices.