A-MEM
AGI ResearchA-MEM is a Python memory library for LLM agents that stores each memory as a note with model-generated keywords, tags and context, links it to related notes in ChromaDB, and revises the context of older notes as new ones arrive, after the Zettelkasten method. It implements the A-MEM paper and installs from source. The AGI Research group publishes it with the paper's authors.
Read against agiresearch/A-mem. The first author's WujiangXu/A-mem-sys carries the same package with added model backends, and the paper's experiments live in WujiangXu/AgenticMemory.
Openness
5 high confidence- license
- MIT (the agentic_memory package)
- source
- public
- core features withheld
- no — the 19-path tree at ceffb86 holds the agentic_memory package, tests, an example and the root MIT LICENSE, with no ee/, enterprise/, commercial/ or proprietary/ path; a research group's release with nothing sold beside it
A-MEM is published under the MIT license, and the whole memory system installs from the repository and runs against the user's own OpenAI or Ollama models. It is a research group's release of a paper's method, with no hosted service or paid edition beside it.
- https://raw.githubusercontent.com/agiresearch/A-mem/main/LICENSE recorded 2026-11-20
The LICENSE is the MIT License, "Copyright (c) 2025 AGI Research".
- https://raw.githubusercontent.com/agiresearch/A-mem/main/README.md recorded 2026-11-20
The README says "This repository provides a memory system to facilitate agent construction", points paper reproduction to WujiangXu/AgenticMemory, installs with "pip install ." and lists OpenAI and Ollama backends; "This project is licensed under the MIT License."
- https://raw.githubusercontent.com/WujiangXu/A-mem-sys/main/README.md recorded 2026-11-20
The A-mem-sys README describes the same Agentic Memory system and tells readers to "git clone https://github.com/agiresearch/A-mem.git".
Adoption
2 low confidenceA-MEM installs from source and has no package on PyPI under its own name, so GitHub stars are the only signal. A star is not a use, and stars on a paper's companion repository mostly record interest in the research.
- https://api.github.com/search/repositories?q=repo:agiresearch/A-mem recorded 2026-11-20
agiresearch/A-mem: 1,216 stars, license MIT, not archived, last pushed 2025-12-12.
- https://pypi.org/pypi/agentic-memory/json recorded 2026-11-20
The agentic-memory package on PyPI belongs to ARYAN2302/ContextOS, a different project, so it is not this repository's release.
Capability
2 medium confidenceA-MEM's distinguishing step is evolution: each new note makes a model revisit its neighbors, link them and rewrite their context. It does not decide what to remember from a conversation, though, storing whatever text the agent hands it, and nothing in it expires or fades over time.
- https://raw.githubusercontent.com/agiresearch/A-mem/main/README.md recorded 2026-11-20
"When a new memory is added to the system: Generates comprehensive notes with structured attributes ... Analyzes historical memories for relevant connections ... Enables dynamic memory evolution and updates"; the API adds notes with add_note and searches with search_agentic.
Verified 2026-11-20, except some scores, still awaiting re-confirmation