thedotmack / claude-mem
Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions.
View thedotmack/claude-memAn LLM call is stateless: everything the model knows about your session has to travel inside the prompt. Agent memory is the layer in between — it decides what gets extracted from a conversation, where that state is stored, and what is pulled back into the context window on the next turn. The projects on this board attack that problem from different angles: embedded stores that run in the same process as the agent, graph memory that keeps entities and relations, vector-backed recall, and systems that summarise a transcript instead of storing it verbatim. Entries come from GitHub topic pages for agent memory, long-term memory, LLM memory and memory management, ranked by stars, so a heavyweight framework and a small single-file library land in the same list. Every card carries language, stars and forks; every page adds description, license, last push date, README and related projects. Stars measure attention, not maintenance — read the activity dates before you give an agent the keys to real user data.
Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions.
View thedotmack/claude-memA Simple and Universal Swarm Intelligence Engine, Predicting Anything. 简洁通用的群体智能引擎,预测万物
View 666ghj/MiroFishThe Memory Layer for AI Agents - Drop-in memory infrastructure for AI agents and apps. Context that persists. Built for production.
View mem0ai/mem0《深入理解 AI Agent:设计原理与工程实践》(李博杰 著)开源主仓库:全书正文、编译版 PDF 与按章配套代码
View bojieli/ai-agent-bookTiDB is built for agentic workloads that grow unpredictably, with ACID guarantees and native support for transactions, analytics, and vector search. No data silos. No noisy neighbors.
View pingcap/tidbSelf-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.
View volcengine/OpenVikingCognee is the open-source AI memory platform for agents. Give your AI agents persistent long-term memory across sessions with a self-hosted knowledge graph engine.
View topoteretes/cogneeMemory and context engine + app that is extremely fast, scalable, and can be run fully locally. The Memory API for the AI era.
View supermemoryai/supermemoryTencentDB Agent Memory is a team-level memory hub for AI Agents — turning conversations, docs, and code into four reusable memory assets (Chat Memory, Skill, LLM-Wiki, Code-Graph) that are governed
View TencentCloud/TencentDB-Agent-MemoryMemori is agent-native memory infrastructure. A LLM-agnostic layer that turns agent execution and conversation into structured, persistent state for production systems.
View MemoriLabs/MemoriGraph-Native Infrastructure for Context and Accountable AI Systems
View semantica-agi/semanticaOne portable memory layer for every AI agent: local-first, Markdown-native, user-owned, and self-evolving across apps, tools, and workflows.
View EverMind-AI/EverOSSelf-evolving memory OS for LLM & AI Agents: ultra-persistent memory, hybrid-retrieval, and cross-task skill reuse, with 35.24% token savings and DeepSeek Harness support.
View MemTensor/MemOSLightweight real-time memory management application to monitor and clean system memory on your computer.
View henrypp/memreductThis free RAM cleaner uses native Windows features to optimize memory areas. It's a compact, portable, and smart application.
View IgorMundstein/WinMemoryCleanerYour agents don’t hand off the work. Kungfu keeps the same Work moving across Codex, Claude, OpenCode, and your own execution surface.
View kungfu-systems/kungfuLocal persistent memory store for LLM applications including claude desktop, github copilot, codex, antigravity, etc.
View CaviraOSS/LongMemoryA bio-inspired cognitive memory engine — a new paradigm for Graph RAG.
View FlowElement-xinliuyuansu/m_flowAwesome list for AI agent harness engineering: tools, patterns, evals, memory, MCP, permissions, observability, and orchestration.
View ai-boost/awesome-harness-engineeringA smarter, self-hosted AI assistant — multi-user, multi-agent.
View TencentCloud/OctopThe Boehm-Demers-Weiser conservative C/C++ Garbage Collector (bdwgc, also known as bdw-gc, boehm-gc, libgc)
View bdwgc/bdwgcEasy to integrate Vulkan memory allocation library
View GPUOpen-LibrariesAndSDKs/VulkanMemoryAllocatorMirix is a multi-agent personal assistant designed to track on-screen activities and answer user questions intelligently.
View Mirix-AI/MIRIXOP Vault ChatGPT: Give ChatGPT long-term memory using the OP Stack (OpenAI + Pinecone Vector Database). Upload your own custom knowledge base files (PDF, txt, epub, etc) using a simple React frontend.
View pashpashpash/vault-aiAmazon Bedrock Agentcore accelerates AI agents into production with the scale, reliability, and security, critical to real-world deployment.
View awslabs/agentcore-samplesStateful agents that are like people, with memory, identity, and the ability to learn and adapt
View letta-ai/letta-codeNeo.mjs is a self-evolving software organism: a professional end-to-end AI engineering team whose cross-model swarm inhabits live apps via Neural Link, Active Hybrid GraphRAG, DreamService
View neomjs/neoUniversal memory layer for AI Agents. It provides scalable, extensible, and interoperable memory storage and retrieval to streamline AI agent state management for next-generation autonomous systems.
View MemMachine/MemMachineUser Profile-Based Long-Term Memory for AI Chatbot Applications.
View memodb-io/memobaseA persistent, unified memory layer for all your AI agents (e.g. Claude Code, Codex, DSH), backed by Markdown and Milvus.
View zilliztech/memsearch🧠 Make your agents learn from experience. Now available as a hosted solution at kayba.ai
View kayba-ai/agentic-context-engineBitterbot - a mesh of agents that turns shared experience into collective capability.
View Bitterbot-AI/bitterbot-desktopPortable project memory across Claude Code, Codex and OpenCode, plus token accounting measured from harness transcripts. Local file I/O, no API calls, no telemetry.
View cytostack/openwolf📰 Must-read papers and blogs on LLM based Long Context Modeling 🔥
View Xnhyacinth/Awesome-LLM-Long-Context-ModelingUnbounded context. Memory that manages itself. One session, for life. The hippocampus for coding agents, part of CortexKit.
View cortexkit/magic-contextThe open-source RAG platform: built-in citations, deep research, 22+ file formats, partitions, MCP server, and more.
View agentset-ai/agentsetInfrastructure for the next generation of voice agents, designed to provide universal memory. It is divided into a left brain and a right brain, storing information and emotions respectively
View xzf-thu/VoiceMemUnofficial implementation of Titans, SOTA memory for transformers, in Pytorch
View lucidrains/titans-pytorchOpen-source persistent memory for AI agent pipelines (LangGraph, CrewAI, AutoGen) and Claude. REST API + knowledge graph + autonomous consolidation.
View doobidoo/mcp-memory-service🍙 A personal AI agent & local memory hub for all AI agents, gives every AI one shared, fully controlled memory and persistent context — all AI remember the same you.
View MemTensor/memmy-agentFind memory issues & leaks in your iOS app without instruments
View tapwork/HeapInspector-for-iOSA memory allocator that automatically reduces the memory footprint of C/C++ applications.
View plasma-umass/MeshHivemind turns your traces into reusable skills across agents
View activeloopai/hivemindTeam memory for engineers and their AI agents. Lives in your repo. Shared through Git.
View mex-memory/mex[EMNLP 2025 Oral] MemoryOS is designed to provide a memory operating system for personalized AI agents.
View BAI-LAB/MemoryOSA free, open-source Chrome extension - no ads, no tracking. Based on The Great Suspender, cleaned up and actively maintained.
View gioxx/MarvellousSuspenderResearch of DeepSeek Engram Architecture based on Qwen-3 and Stable Diffusion series.
View AutoArk/TinyEngramA lightweight, rollbackable, and visual Long-Term Memory Server for MCP Agents. Say goodbye to Vector RAG and amnesia.
View Dataojitori/nocturne_memory总结整理linux内核的内存管理的资料,包含论文,文章,视频,以及应用程序的内存泄露,内存池相关
View 0voice/kernel_memory_managementAwesome AI Memory | LLM Memory | A curated knowledge base on AI memory for LLMs and agents, covering long-term memory, reasoning, retrieval, and memory-native system design.
View IAAR-Shanghai/Awesome-AI-Memory[ICLR 2026] LightMem: Lightweight and Efficient Memory-Augmented Generation
View zjunlp/LightMemAgent memory for LLMs: 30 runnable Jupyter notebooks covering conversation buffers, vector stores, knowledge graphs, episodic and semantic memory, MemGPT, Mem0, Letta, Zep, Graphiti
View NirDiamant/Agent_Memory_TechniquesEcho Agent 是一个可自托管、长期运行、持续学习的 AI Agent,面向个人与团队的私有自动化场景。它可以部署在自有服务器上,统一连接模型、工具、记忆、权限与消息入口。内置四层认知记忆、遗忘曲线与矛盾检测机制,能够在跨会话任务中持续沉淀上下文,并保持长期记忆的质量。针对命令执行、文件操作等高风险行为,它提供基于 LLM 的审批与解释机制,为关键操作建立可审计、可追溯的安全边界。原生支持
View fuyuxiang/echo-agentEasy to integrate memory allocation library for Direct3D 12
View GPUOpen-LibrariesAndSDKs/D3D12MemoryAllocatorA simple guide to x86 architecture, assembly, memory management, paging, segmentation, SMM, BIOS....
View Captainarash/The_Holy_Book_of_X86Long-term memory runtime for AI agents — plain Markdown as the source of truth, local ranked retrieval, and an independent sleep-time Manage layer. Claude Code and Codex share one store. No API key.
View tigerless-labs/agent-memoryAgentChat 是一个基于 LLM 的智能体交流平台,内置默认 Agent 并支持用户自定义 Agent。通过多轮对话和任务协作,Agent 可以理解并协助完成复杂任务。项目集成 LangChain、Function Call、MCP 协议、RAG、Memory、HITL、Skill、Milvus 和 ElasticSearch 等技术,实现高效的知识检索与工具调用,使用 FastAPI
View Shy2593666979/AgentChatA cyber brain for your AI. It never forgets a detail, remembers exactly what you said, and learns how you work over time.
View CodeAbra/iai-personal-memory-engineOne memory shared by Claude Code, Codex, Cursor, Copilot CLI, OpenClaw and 28 more coding agents, built from the session history already on disk.
View vshulcz/deja-vuOpen-source coding agents memory. Records issues, attempts, fixes and decisions, then warns your agent before it repeats an approach that already failed.
View riponcm/projectmemOfficial implementation of our NeurIPS 2023 paper "Augmenting Language Models with Long-Term Memory".
View Victorwz/LongMemOpen-source cross-agent memory layer for coding agents via MCP. Compatible with Claude Code, Codex, Cursor, Windsurf, Gemini CLI, Antigravity, OpenClaw, Hermes Agent, Oh-my-Pi, Pi, Copilot, Kiro
View AVIDS2/memorixRedis memory profiler to find the RAM bottlenecks throw scaning key space in real time and aggregate RAM usage statistic by patterns.
View gamenet/redis-memory-analyzerTurns corrections into Preferences, Project-specific skills, and Shared skills for Claude Code, Codex, and OpenCode.
View ReflexioAI/claude-smartAn Innovative Agent Framework Driven by KG Engine
View codefuse-ai/CodeFuse-muAgentNVIDIA RMM is a library for allocating and managing GPU memory in C++ and Python.
View rapidsai/rmm司南:个性化 AI 任务总控 Skills 系统 /COMPASS: Personal Alignment Skills OS for AI Agents
View dongshuyan/compass-skillsGive your AI agents persistent, collective memory — with deduplicating absorb, supersession lineage, semantic search, and a graph UI. Speaks MCP.
View agentic-box/memoraAI agents with graph based reasoning memory, scaffolded in seconds
View neo4j-labs/create-context-graphOpen-source AI companions, desktop pets, long-term memory & proactive chat. 人机恋开源项目大全。让你的家机能够脱离官端自主存在,拥有记忆和主动性。
View DasterProkio/awesome-ai-companionGit-native persistent memory for AI coding agents. Implements Google OKF v0.2 with sub-300µs in-memory BM25 search, embedded MCP server, and progressive disclosure.
View okf-memory/okf-agent-memoryEvidence-first local memory for AI agents with temporal versions, admission policies, citations, explainable recall, MCP, and audit tooling.
View caspianmoon/memoripyThe local-first LLM Wiki: open-source knowledge graph builder, RAG knowledge base, and agent memory store. Built on Andrej Karpathy's pattern.
View swarmclawai/swarmvaultOpen-source self-hosted AI agent runtime and multi-agent framework for autonomous agent swarms. Agent memory, MCP tools, schedules, delegation, and 23+ LLM providers (Claude, GPT, Gemini, OpenRouter
View swarmclawai/swarmclawAutonomous self-evolving agents. Vision-grounded layered memory and self-written skills for LLM agents that operate your computer.
View jmerelnyc/Photo-agentsTypeScript AI agent framework: cognitive memory, runtime tool forging, multi-agent orchestration, 11 LLM providers.
View framerslab/agentosGive your AI a real, persistent memory. The open-source system plus templates that turn an Obsidian vault into your AI's working memory. No vector database, just markdown.
View jaredrhod/ai-memory-vaultOpen-source infrastructure that turns scattered SKILL.md files into curated, retrieval-ready agent-skill corpora—with retrieval and evaluation tooling included.
View EverMind-AI/SkillCorpusCurated systems, benchmarks, and papers etc. on memory for LLMs/MLLMs --- long-term context, retrieval, and reasoning.
View TeleAI-UAGI/Awesome-Agent-MemoryOS-level autonomous AI agent with long-term memory, multi-agent coordination, Titan Chronos scheduler & Moltbot Social Core
View Arvincreator/project-golemANOLISA (Agentic Nexus Operating Layer & Interface System Architecture) | Agentic OS with runtime, security, observability, and Tokenless response compression for lower token usage and cost.
View agentic-os-org/ANOLISACognitive Deterministic Memory Security OS for Agentic AI that reaches backward through time to find the quiet change behind today's failure, not the lookalike.
View samvallad33/vestigeA language model has no memory of its own. Every call starts blank, so anything an agent needs to know — a user preference, a decision taken three sessions ago, the contents of a repository it just read — has to be reconstructed and placed into the prompt. Agent memory is the machinery that does the reconstruction. It is usually three jobs glued together: extraction (deciding which part of a conversation is worth keeping), storage (a vector index, a graph, a relational table, or plain files on disk) and retrieval (choosing what goes back into the context window on the next call). Almost every project on this board is an opinionated answer to one question: where does state live, and how is it fetched.
The distinction that matters in production is not the short-term versus long-term label, but the prompt-token budget. A system that stores every turn and replays all of it drowns the context window and pays for it on every request; a system that retrieves three relevant facts costs almost nothing. That trade-off — how much to remember versus how much to re-send — separates a weekend demo from something you can run for a month.
Long-term memory is whatever survives the end of a session: facts about a person, decisions taken on a project, the state of a task. Three shapes dominate the projects below. Summarisation compresses the running transcript so recent turns stay cheap to send. Vector recall embeds each memory as text plus an embedding and fetches the nearest by similarity — quick to build, weak at questions that need structure. Graph memory extracts people, projects and relations, then walks them at query time; it answers questions like which library did we drop and why far better than similarity search, at the cost of an extraction step you must maintain.
The three are not exclusive. The stronger systems here write to two stores at once — a cheap summary for the live conversation, and a structured, searchable store for everything older.
On-device memory means the memory layer runs beside the model — same process, same container, same laptop — instead of posting user data to a hosted service. Two families dominate. Embedded stores are essentially a database file with vector search bolted on: no server, no network, no per-query bill. Self-hosted servers buy you a retention policy you can actually explain to a user and a deletion path that does not depend on someone else’s API. If data residency or offline operation is the requirement, start with the embedded tools — they are the ones that keep working with the network cable unplugged.
What is an AI agent memory system? The component that persists information between model calls and decides what goes back into the context window: normally an extraction step, a store, and a retrieval policy.
Does agent memory need a vector database? No. Vector search is one retrieval strategy and the easiest to start with. Graph stores, SQLite tables and flat files are all used in production, especially when the questions you ask are structural rather than semantic.
What is on-device memory for AI agents? A memory store running on the same machine or in the same process as the agent, so nothing is sent to a third party and recall keeps working offline. Embedded databases and local vector indexes are the usual building blocks.
Is agent memory the same as fine-tuning? No. Fine-tuning changes the weights and is expensive to update. Memory changes what is in the prompt, which means you can edit, delete and audit it per user — usually exactly what privacy rules demand.
How do I stop memory from growing forever? Give it a write policy and a decay rule: score memories as they are written, expire the low-value ones, and cap the retrieved set. Every mature project on this board exposes some version of that policy.