A-Mem: Agentic Memory for LLM Agents

Researchers Wujiang Xu and colleagues posted A-MEM, an agentic memory system for LLM agents, to arXiv on 17 February 2025, with the paper last revised 8 October 2025 as version v11. A-MEM follows the Zettelkasten method, generating structured notes with contextual descriptions, keywords and tags for each new memory, then dynamically linking related historical memories and letting new memories trigger updates to existing ones. Experiments across six foundation models showed superior improvement over existing state-of-the-art baselines, with evaluation code at github.com/WujiangXu/A-mem and the system code at github.com/WujiangXu/A-mem-sys.

Computer Science Computation and Language Submitted on 17 Feb 2025 v1 https://arxiv.org/abs/2502.12110v1 , last revised 8 Oct 2025 this version, v11 Title:A-MEM: Agentic Memory for LLM Agents View PDF https://arxiv.org/pdf/2502.12110 HTML experimental https://arxiv.org/html/2502.12110v11 Abstract:While large language model LLM agents can effectively use external tools for complex real-world tasks, they require memory systems to leverage historical experiences. Current memory systems enable basic storage and retrieval but lack sophisticated memory organization, despite recent attempts to incorporate graph databases. Moreover, these systems' fixed operations and structures limit their adaptability across diverse tasks. To address this limitation, this paper proposes a novel agentic memory system for LLM agents that can dynamically organize memories in an agentic way. Following the basic principles of the Zettelkasten method, we designed our memory system to create interconnected knowledge networks through dynamic indexing and linking. When a new memory is added, we generate a comprehensive note containing multiple structured attributes, including contextual descriptions, keywords, and tags. The system then analyzes historical memories to identify relevant connections, establishing links where meaningful similarities exist. Additionally, this process enables memory evolution - as new memories are integrated, they can trigger updates to the contextual representations and attributes of existing historical memories, allowing the memory network to continuously refine its understanding. Our approach combines the structured organization principles of Zettelkasten with the flexibility of agent-driven decision making, allowing for more adaptive and context-aware memory management. Empirical experiments on six foundation models show superior improvement against existing SOTA baselines. The source code for evaluating performance is available at this https URL https://github.com/WujiangXu/A-mem , while the source code of the agentic memory system is available at this https URL https://github.com/WujiangXu/A-mem-sys . Submission history From: Wujiang Xu view email https://arxiv.org/show-email/8861f1bb/2502.12110 Mon, 17 Feb 2025 18:36:14 UTC 603 KB \ v1\ https://arxiv.org/abs/2502.12110v1 Mon, 3 Mar 2025 04:14:02 UTC 603 KB \ v2\ https://arxiv.org/abs/2502.12110v2 Tue, 4 Mar 2025 15:09:10 UTC 603 KB \ v3\ https://arxiv.org/abs/2502.12110v3 Mon, 14 Apr 2025 15:21:49 UTC 603 KB \ v4\ https://arxiv.org/abs/2502.12110v4 Fri, 18 Apr 2025 17:26:57 UTC 603 KB \ v5\ https://arxiv.org/abs/2502.12110v5 Sun, 11 May 2025 18:10:25 UTC 2,620 KB \ v6\ https://arxiv.org/abs/2502.12110v6 Wed, 21 May 2025 05:16:32 UTC 2,629 KB \ v7\ https://arxiv.org/abs/2502.12110v7 Tue, 27 May 2025 02:44:13 UTC 1,002 KB \ v8\ https://arxiv.org/abs/2502.12110v8 Mon, 2 Jun 2025 22:21:21 UTC 995 KB \ v9\ https://arxiv.org/abs/2502.12110v9 Tue, 15 Jul 2025 00:44:52 UTC 600 KB \ v10\ https://arxiv.org/abs/2502.12110v10 v11 Wed, 8 Oct 2025 01:46:37 UTC 607 KB References & Citations Loading... Bibliographic and Citation Tools Bibliographic Explorer What is the Explorer? https://info.arxiv.org/labs/showcase.html arxiv-bibliographic-explorer Connected Papers What is Connected Papers? https://www.connectedpapers.com/about Litmaps What is Litmaps? https://www.litmaps.co/ scite Smart Citations What are Smart Citations? https://www.scite.ai/ Code, Data and Media Associated with this Article alphaXiv What is alphaXiv? https://alphaxiv.org/ CatalyzeX Code Finder for Papers What is CatalyzeX? https://www.catalyzex.com DagsHub What is DagsHub? https://dagshub.com/ Gotit.pub What is GotitPub? http://gotit.pub/faq Hugging Face What is Huggingface? https://huggingface.co/huggingface ScienceCast What is ScienceCast? https://sciencecast.org/welcome Demos Recommenders and Search Tools Influence Flower What are Influence Flowers? https://influencemap.cmlab.dev/ CORE Recommender What is CORE? https://core.ac.uk/services/recommender arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs https://info.arxiv.org/labs/index.html .