MIRIX
Open-source multi-agent memory system: six memory types (core, episodic, semantic, procedural, resource, knowledge vault) managed by dedicated agents, with optional screen capture.
Tags
Description
Run the backend and dashboard with Docker Compose, create an API key in the dashboard and use the Python client (pip install mirix-client) to give an assistant long-term memory built from conversations and, optionally, continuous screen observation. Long-term data is stored locally; search uses PostgreSQL-native BM25 full-text plus vector similarity. Inputs can be text, images, voice and screen captures. Described in arXiv 2507.07957.
Related
- Mem0: Memory layer for AI agents and apps: drop-in persistent memory that lets agents learn from past interactions.
- Letta: Platform for stateful agents with memory, identity and the ability to learn over time (formerly MemGPT); terminal UI, App Server, desktop app and Agent SDK.
- Zep: Unified context layer that combines business data, documents and conversations into governed context and memory for agents.
- MemOS: Open-source memory operating system for LLMs and agents with a unified API for adding, retrieving, editing and deleting graph-structured, multi-modal memory.
- Cognee: Open-source AI memory platform for agents that turns documents and conversations into graph, vector and relational memory.
Sources
Machine-readable
- JSON
- Markdown
- Source file on GitHub (edit via pull request)