Agent Memory

Agent Memory is the set of mechanisms that let an AI agent retain and reuse information beyond a single context window — across turns, sessions, tools, and tasks — instead of starting blank each time.

It usually spans short-term working memory (the active context) and longer-term stores that are written, summarized, and retrieved on demand. The design questions are what to remember, how to compress it, and when to fetch it back, closely related to RAG and Context Engineering.

For builders it is what makes multi-step Agents reliable over long horizons; without it an agent repeats work and loses track of its goal. The failure mode is unbounded memory that floods the window with stale detail, degrading the very reasoning it was meant to support.

Topic: Agents Related: Agents Context Engineering RAG

Recent Updates

• 2026-09-05: Local Zoom assistant, part 2: how meeting graph becomes memory and why you can trust it (breakingnewsofficial) · habr.com — Obsidian · Zoom
• 2026-09-05: OKF Agent Memory: Git-native persistent memory for AI coding agents (breakingnewsofficial) · github.com — Google · Gemma · Qwen · LLaMA
• 2026-09-06: Embodied AI Startups Adopt In-Context Learning as New Scaling Path (breakingnewsofficial) · qbitai.com — Skild AI Generalist AI COCO Matrix Fourier Intelligence Gao Yuxiang Li Fei-Fei Jim Fan Yuke Zhu S1 GEN-1.5 RoboTTT GPT-3
• 2026-09-07: KVMem: Virtualizing Million-Token Agent Workspaces on a Consumer GPU (breakingnewsofficial) · arxiv.org — Qwen3.8-27B · Qwen3.6-27B
• 2026-09-07: Rhythms of Work: Multi-Scale Interpretation of Human Behavioral Traces for Workplace Agents (breakingnewsofficial) · arxiv.org
• 2026-09-07: La Agente ‘Optima: Towards Agentic Self-Driving Laboratories (breakingnewsofficial) · arxiv.org
• 2026-09-07: Engrim: Local-first SQLite memory engine for cross-model AI coding environments (breakingnewsofficial) · github.com — Google · Anthropic · OpenAI Gemini 3.8 · Claude 3.7 Sonnet · GPT-4o
• 2026-09-08: TradingAgents: Open-source multi-agent framework for AI-powered financial analysis (breakingnewsofficial) · github.com — OpenAI · Anthropic · Google · Groq · Mistral · AWS · Microsoft · NVIDIA · Kimi · MiniMax · DeepSeek · Zhipu · Alibaba Alpha Vantage FRED · Polymarket StockTwits · Reddit · GPT-5.6 · GPT 5.5 · GPT-5.4 · Claude Sonnet 5 · Claude 4.6 · Fable 5 · Gemini 3.1 Grok 4.x · Qwen · GLM
• 2026-09-09: Memory in Deep Time-Series Models: A Unified Framework for Retention and Access (breakingnewsofficial) · arxiv.org
• 2026-09-09: EdgeMem: LLM-Free Agent Memory Construction and Retrieval via Evidence-Preserving Multi-Anchor Hypergraph (breakingnewsofficial) · arxiv.org
• 2026-09-09: Multi-turn LLM Degradation: How Assistant-Generated History Shapes Downstream Behavior (breakingnewsofficial) · arxiv.org
• 2026-09-09: SCAFFOLD: Self-Improving Web Agents via Recursive Parametric Skill Abstraction (breakingnewsofficial) · arxiv.org
• 2026-09-09: AutoFyn: Expert Iteration for Long-Horizon Agents via Persistent State Adaptation (breakingnewsofficial) · arxiv.org — AutoFyn Next.js MetaMask · pnpm · Warp · LiteLLM · Langflow · Open WebUI
• 2026-09-09: MERIT: Cost-Aware Evaluation of Memory in Tool-Using LLM Agents (breakingnewsofficial) · arxiv.org — OpenAI · Anthropic · GPT-4.1-mini · GPT-4.1 · Claude Haiku 4.5 · Claude Sonnet 5
• 2026-09-09: OtoDock: Self-hosted agent platform with persistent memory and team coordination (breakingnewsofficial) · github.com — Anthropic · OpenAI · Twilio
• 2026-09-10: Fortunate Recall: Ontology-Driven Memory Lifecycle Management for Persistent Coherence in LLMs (breakingnewsofficial) · arxiv.org — kimi-k2.5
• 2026-09-10: Separating Agent Storage from Usage: RD-Forget Framework for Persistent Memory Management (breakingnewsofficial) · arxiv.org
• 2026-09-10: Kernel-Managed Shared Memory for Multi-Agent System Personalization (breakingnewsofficial) · arxiv.org — OpenAI · Meta · Alibaba · Mem0 · GPT-4o · Llama-3.1:8B · Qwen 2.5 7B
• 2026-09-10: Research Attention Prediction: Evaluating LLM Agents’ Ability to Forecast AI Research Trends (breakingnewsofficial) · arxiv.org — OpenAI · Alibaba · GPT 5.5 · Qwen3-4B
• 2026-09-10: VLX-VR: An Agentic-Aware Video Reasoning Model (breakingnewsofficial) · arxiv.org — VLX-VR
• 2026-09-10: AgentAudit: An Open, Extensible Framework for Full-Lifecycle Trust Evaluation of AI Agents (breakingnewsofficial) · arxiv.org — OpenAI · Anthropic · Google · Meta Sarvam AI · GPT-5 · Claude Sonnet 5 Sarvam 105B · Llama-3.3-70b · Gemini 2.5 Flash
• 2026-09-10: Procedural Memory Under Change: Reuse and Interference in Controlled Web Tasks (breakingnewsofficial) · arxiv.org — Qwen3-8B
• 2026-09-10: ROAM: Robust Organization of Atomic Memories for Agents through Semantic Relations (breakingnewsofficial) · arxiv.org
• 2026-09-10: PRAGMA: Evaluating Personalized Guidance with Memory Alignment in Lifelong Conversations (breakingnewsofficial) · arxiv.org
• 2026-09-11: Auto-RecSys: Harnessing Autonomous Research Agents for Industry-Scale Recommender Systems (breakingnewsofficial) · arxiv.org

FAQ

What is Agent Memory?

Agent Memory is the set of mechanisms that lets AI agents persist useful context across turns, sessions, tools, and tasks. GROUNDING tracks memory stores, compression, retrieval boundaries, and failure modes for long-running agent workflows.

Which topic does Agent Memory belong to?

On the GROUNDING radar, Agent Memory is grouped under the Agents topic.

Which concepts are related to Agent Memory?

Related concepts tracked by the radar include Agents, Context Engineering, RAG.