[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"pack-detail-agent-memory-comparison-fr":3,"seo:pack:agent-memory-comparison:fr":101},{"code":4,"message":5,"data":6},200,"操作成功",{"pack":7},{"slug":8,"icon":9,"tone":10,"status":11,"status_label":12,"title":13,"description":14,"items":15,"install_cmd":100},"agent-memory-comparison","🧬","#F59E0B","new","Nouveau · cette semaine","Comparatif Mémoire pour Agents : Mem0 vs Letta vs Graphiti vs Zep","Dix couches mémoire comparées : Mem0, Letta (MemGPT), Graphiti, Zep, Cognee, Memvid, Memori, Engram, Statewave + la Awesome list. Du léger au graphe au managé, matrice de décision incluse. Opinions tranchées.",[16,28,36,46,54,61,69,77,86,93],{"id":17,"uuid":18,"slug":19,"title":20,"description":21,"author_name":22,"view_count":23,"vote_count":24,"lang_type":25,"type":26,"type_label":27},703,"96da1f40-1823-4d87-a84f-7d8269edeb24","mem0-memory-layer-ai-applications-96da1f40","Mem0 — Memory Layer for AI Applications","Add persistent, personalized memory to AI agents and assistants. Mem0 stores user preferences, past interactions, and learned context across sessions.","Mem0",713,0,"en","skill","Skill",{"id":29,"uuid":30,"slug":31,"title":32,"description":33,"author_name":34,"view_count":35,"vote_count":24,"lang_type":25,"type":26,"type_label":27},302,"4ddbd0e0-80df-41ee-9593-982beeb075b3","letta-stateful-ai-agents-memory-4ddbd0e0","Letta — Stateful AI Agents with Memory","Letta builds stateful AI agents that learn and self-improve with advanced memory. 21.8K+ stars. CLI, Python\u002FTS SDKs, skills, subagents. Apache 2.0.","Skill Factory",162,{"id":37,"uuid":38,"slug":39,"title":40,"description":41,"author_name":42,"view_count":43,"vote_count":24,"lang_type":25,"type":44,"type_label":45},211,"34ea44af-b30f-4129-b6fd-a2cb4366ab4b","graphiti-real-time-knowledge-graphs-ai-agents-34ea44af","Graphiti — Real-Time Knowledge Graphs for AI Agents","Build real-time knowledge graphs for AI agents by Zep. Temporal awareness, entity extraction, community detection, and hybrid search. Production-ready. 24K+ stars.","TokRepo精选",227,"script","Script",{"id":47,"uuid":48,"slug":49,"title":50,"description":51,"author_name":52,"view_count":53,"vote_count":24,"lang_type":25,"type":26,"type_label":27},743,"ffde39a9-bd16-4c41-8168-aacfb05b7622","zep-long-term-memory-ai-agents-assistants-ffde39a9","Zep — Long-Term Memory for AI Agents and Assistants","Production memory layer for AI assistants. Zep stores conversation history, extracts facts, builds knowledge graphs, and provides temporal-aware retrieval for LLMs.","MCP Hub",195,{"id":55,"uuid":56,"slug":57,"title":58,"description":59,"author_name":34,"view_count":60,"vote_count":24,"lang_type":25,"type":26,"type_label":27},334,"b6ad223f-51da-4179-863c-5f9b3c7d08ef","cognee-memory-engine-ai-agents-b6ad223f","Cognee — Memory Engine for AI Agents","Cognee adds persistent structured memory to any AI agent in 6 lines of code. 14.8K+ stars. Knowledge graphs, vector stores, LLM integration. Apache 2.0.",146,{"id":62,"uuid":63,"slug":64,"title":65,"description":66,"author_name":67,"view_count":68,"vote_count":24,"lang_type":25,"type":26,"type_label":27},2631,"c6f6a833-47b4-11f1-9bc6-00163e2b0d79","memvid-serverless-memory-layer-ai-agents-c6f6a833","Memvid — Serverless Memory Layer for AI Agents","An open-source memory system that replaces complex RAG pipelines with a single-file, serverless memory layer providing instant retrieval and long-term storage for AI agents.","Script Depot",138,{"id":70,"uuid":71,"slug":72,"title":73,"description":74,"author_name":75,"view_count":76,"vote_count":24,"lang_type":25,"type":26,"type_label":27},3541,"7fcb5dcd-8666-5767-a3af-208f83343287","memori-agent-native-memory-infrastructure","Memori — Agent-Native Memory Infrastructure","Memori is an Apache-2.0 memory layer that captures what agents do (not just say) and plugs into existing stacks via Python\u002FNode SDKs and a cloud option.","AI Open Source",76,{"id":78,"uuid":79,"slug":80,"title":81,"description":82,"author_name":52,"view_count":83,"vote_count":24,"lang_type":25,"type":84,"type_label":85},600,"f5bac4b3-caff-48a5-8b6a-48de91fec697","engram-persistent-memory-system-ai-agents-f5bac4b3","Engram — Persistent Memory System for AI Agents","Agent-agnostic persistent memory system with SQLite full-text search. Ships as MCP server, HTTP API, CLI, and TUI. Gives any AI coding agent long-term memory across sessions. 2,300+ stars.",174,"mcp","MCP",{"id":87,"uuid":88,"slug":89,"title":90,"description":91,"author_name":75,"view_count":92,"vote_count":24,"lang_type":25,"type":26,"type_label":27},3367,"e900097c-4946-5899-829b-0b9db4c2adf7","statewave-memory-runtime-for-ai-agents-api-sdks","Statewave — Memory Runtime for AI Agents (API + SDKs)","Statewave is a self-hostable memory runtime: ingest episodes, compile memories, do semantic search, and build token-bounded context bundles via REST.",55,{"id":94,"uuid":95,"slug":96,"title":97,"description":98,"author_name":75,"view_count":99,"vote_count":24,"lang_type":25,"type":26,"type_label":27},3318,"be6dfe8e-975e-5ade-9900-72221c32ab40","awesome-agent-memory-long-term-context-index","Awesome Agent Memory — Long-Term Context Index","Awesome Agent Memory curates systems, benchmarks, and papers on long-term context for LLMs\u002FMLLMs—use it to compare approaches and pick tools to try.",87,"tokrepo install pack\u002Fagent-memory-comparison",{"pageType":102,"pageKey":8,"locale":25,"title":103,"metaDescription":104,"h1":105,"tldr":106,"bodyMarkdown":107,"faq":108,"schema":124,"internalLinks":134,"citations":147,"wordCount":160,"generatedAt":161},"pack","Agent Memory Layer Comparison: Mem0 vs Letta vs Graphiti vs Zep (2026)","Ten agent memory layers benchmarked: Mem0, Letta (MemGPT), Graphiti, Zep, Cognee, Memvid, Memori, Engram, Statewave + Awesome index. Install order, tradeoff matrix, and the one question that picks for you.","Agent Memory: The Honest Comparison","There is no best memory layer — only the right one for your shape of problem. This pack lines up ten options from lightweight library to hosted service to graph engine, in install order, with a decision matrix that tells you which to pick by answering one question: how much state does a session actually need to survive?","## Why this pack exists\n\nOur earlier `agent-memory-layer` pack covered the three canonical libraries — Mem0, Zep, Cognee — and four design patterns. Useful, but every week someone on Twitter posts \"which memory layer should I use for X\" and gets ten contradictory answers. This pack is the honest tradeoff guide. Ten options, opinionated install order, decision matrix at the bottom. Read it once, pick, ship.\n\nThe field has fragmented since mid-2025. We now have library-shaped solutions (Mem0, Memori), service-shaped solutions (Zep, Statewave), graph-shaped solutions (Graphiti, Cognee), stateful-runtime solutions (Letta, Engram), and weird outliers (Memvid encodes memory as MP4 frames). They are not interchangeable. Picking the wrong one buries a year of work.\n\n## Install order — lightweight to heavy\n\n| # | Asset | Shape | When it wins |\n|---|---|---|---|\n| 1 | **Mem0** | Python\u002FNode library | Single-user chatbot, weekend prototype, you want one import |\n| 2 | **Letta** (was MemGPT) | Stateful runtime | Long autonomous agents that need a working-memory + archival split |\n| 3 | **Graphiti** | Temporal graph | Relationships matter and timestamps matter (compliance, audit) |\n| 4 | **Zep** | Hosted service | Multi-tenant SaaS, need ops support, audit trails |\n| 5 | **Cognee** | Graph + vector hybrid library | Research \u002F codebase \u002F domain with strong entity structure |\n| 6 | **Memvid** | Serverless \u002F file-based | Edge agents, no infra, encode memory in a single artifact |\n| 7 | **Memori** | Agent-native infrastructure | Greenfield agent platform, need primitives not products |\n| 8 | **Engram** | Persistent memory system | Self-hosted, ops-light, single-tenant |\n| 9 | **Statewave** | Memory runtime + SDK | Polyglot fleet (Python + TS + Go) sharing one memory |\n| 10 | **Awesome Agent Memory** | Index \u002F taxonomy | You want to evaluate the long tail yourself |\n\nThe order is deliberate: start with Mem0 because the time-to-first-memory is under five minutes. If your agent goes autonomous (multi-step, multi-hour), graduate to Letta — it gives you the explicit two-tier memory split that MemGPT introduced. If your domain is fundamentally relational (who knows whom, what happened when), drop the vector store and start with Graphiti. Zep is the production landing pad for the first two. Cognee is the bet for research-heavy work. Everything below row 5 is a niche pick — read carefully before adopting.\n\n## The decision matrix\n\nAsk yourself **one question**: when this user comes back next month, what does the agent need to remember?\n\n- **\"Their preferences and last few sessions.\"** → Mem0. Stop reading.\n- **\"A multi-hour autonomous task in progress.\"** → Letta. Working + archival memory is the whole point.\n- **\"Who said what, when, to whom.\"** → Graphiti or Zep (Zep ships Graphiti under the hood). Temporal graphs are not optional here.\n- **\"A 6-month corpus of decisions tied to entities.\"** → Cognee or Zep. Vector-only will miss the joins.\n- **\"Whatever — I'll figure it out as I scale.\"** → Mem0 today, plan a Zep migration when you hit 1k users.\n- **\"I cannot run a server.\"** → Memvid or Mem0 with SQLite backend. Hosted Zep cloud is also fine if budget allows.\n- **\"My agent runs in three languages.\"** → Statewave. The SDK parity is the moat.\n\nThe matrix is not exhaustive but it covers ~80% of real shipping agents. If you don't see your case here, it probably belongs in the **Awesome Agent Memory** list (#10) for evaluation rather than a snap decision.\n\n## Where each one breaks\n\n- **Mem0** breaks on multi-tenant audit. You can bolt on user namespaces but there's no first-class audit log. Fine until your first SOC2 question.\n- **Letta** breaks on cost. The two-tier memory means every step does a recall round-trip; for high-frequency agents this doubles your token bill compared to stateless prompts. Worth it for autonomy, not for chat.\n- **Graphiti** breaks on schema drift. If your entity types change weekly during product-market fit, you'll be rewriting graph schemas more than shipping features.\n- **Zep** breaks on self-hosting complexity. Postgres + graph layer + workers; not a Docker run. Use the cloud unless you have a platform team.\n- **Cognee** breaks on opinionation. It assumes you want a knowledge graph; if your domain is loose (general chat), you're paying graph cost for vector value.\n- **Memvid** breaks on size. MP4-frame encoding is clever but the index gets unwieldy past a few GB. Edge-scale only.\n- **Memori \u002F Engram \u002F Statewave** all break on community size. Smaller ecosystems mean fewer integrations, fewer Stack Overflow answers, more bugs you fix yourself.\n\n## Install via TokRepo\n\n```bash\n# Whole pack\ntokrepo install pack\u002Fagent-memory-comparison\n\n# Or the canonical trio\ntokrepo install mem0 letta graphiti\n```\n\nThe CLI drops Claude Code subagent configs into `.claude\u002Fagents\u002F`, Cursor rules into `.cursor\u002Frules\u002F`, AGENTS.md entries for Codex CLI. The pip\u002Fnpm installs are unchanged — TokRepo only wires the AI-tool config so your agent knows the memory layer is available.\n\n## What we left out\n\nWe did **not** include conversation-history hacks (LangChain ConversationBufferMemory and friends). Those aren't memory layers — they're prompt-stuffing wrappers. They work until they don't, and the failure mode is silent context-window overflow. If you find yourself reaching for those, install Mem0 instead. Five minutes, same shape, doesn't blow up at scale.\n\nWe also skipped vector-database-only setups (Pinecone alone, Weaviate alone). Those are RAG infrastructure, not memory. RAG retrieves from a static corpus; memory writes new facts and accumulates. Don't confuse the two.",[109,112,115,118,121],{"q":110,"a":111},"I already use the agent-memory-layer pack — is this a replacement?","No, it's a sibling. The original pack teaches you what a memory layer is and ships the three canonical libraries plus four patterns. This pack assumes you already know that and helps you pick between ten options based on shape of problem. If you're new to agent memory, read agent-memory-layer first. If you've used Mem0 and are wondering whether to switch to Letta or Zep, read this one.",{"q":113,"a":114},"Letta vs MemGPT — are they the same thing?","Letta is the productized successor to MemGPT (the MemGPT paper introduced the two-tier memory architecture; Letta is the open-source framework that implements it as a deployable runtime). If you see MemGPT references in older blog posts, mentally substitute Letta. The architecture is the same: working memory in-context, archival memory in a vector store, with explicit move-between operations.",{"q":116,"a":117},"Is Zep just Graphiti with a UI?","Effectively yes for the temporal-graph half — Zep builds on Graphiti for its memory layer and adds a hosted service, user management, multi-tenant isolation, and an admin UI. If you want self-hosted and ops-friendly, run Graphiti directly. If you want to swipe a credit card and ship, Zep Cloud is the path. The underlying graph semantics are the same.",{"q":119,"a":120},"Why include Memvid? Encoding memory as MP4 sounds insane.","It is, in a fun way. Memvid encodes memory as MP4 frames so the entire memory store is one portable file. For edge agents (CLI tools, IoT, on-device assistants) where you cannot run a server and don't want a SQLite dependency, it's a legitimate option. The constraint is size — past a few gigabytes the index becomes painful. We include it because the constraint shape is unique and there's no other tool in the field that solves the same problem.",{"q":122,"a":123},"What if I outgrow my pick?","Mem0 → Letta is a weekend of refactoring (extract user-facts module, point it at Letta's archival memory). Mem0 → Zep is similar, mostly a config swap. Letta → Zep is harder because you lose Letta's runtime semantics. The riskiest migration is graph (Graphiti \u002F Cognee) to non-graph or vice versa — relationship modeling doesn't survive the trip. Pick the shape right the first time; everything else can be migrated.",{"@context":125,"@type":126,"name":127,"description":128,"numberOfItems":129,"inLanguage":25,"publisher":130},"https:\u002F\u002Fschema.org","CollectionPage","Agent Memory Layer Comparison","Ten agent memory layers benchmarked side-by-side with a decision matrix for picking the right one.",10,{"@type":131,"name":132,"url":133},"Organization","TokRepo","https:\u002F\u002Ftokrepo.com",[135,139,143],{"url":136,"anchor":137,"reason":138},"\u002Fen\u002Fpacks\u002Fagent-memory-layer","Agent Memory Layer (intro pack)","the prerequisite pack for engineers new to agent memory",{"url":140,"anchor":141,"reason":142},"\u002Fen\u002Fpacks\u002Fmulti-agent-frameworks","Multi-Agent Frameworks","memory selection changes when multiple agents share state",{"url":144,"anchor":145,"reason":146},"\u002Fen\u002Fpacks\u002Fml-engineer-rag-eval","RAG + Eval Stack","memory and RAG are complementary; this pack pairs with retrieval evaluation",[148,152,156],{"claim":149,"source_name":150,"source_url":151},"Letta (formerly MemGPT) is a stateful-agent framework that implements the two-tier working\u002Farchival memory architecture from the MemGPT paper","letta-ai\u002Fletta on GitHub","https:\u002F\u002Fgithub.com\u002Fletta-ai\u002Fletta",{"claim":153,"source_name":154,"source_url":155},"Graphiti is a temporal knowledge-graph framework built by the Zep team for AI agent memory","getzep\u002Fgraphiti on GitHub","https:\u002F\u002Fgithub.com\u002Fgetzep\u002Fgraphiti",{"claim":157,"source_name":158,"source_url":159},"Mem0 provides an open-source memory layer that auto-extracts and updates user-specific facts for LLM applications","mem0ai\u002Fmem0 on GitHub","https:\u002F\u002Fgithub.com\u002Fmem0ai\u002Fmem0",920,"2026-05-22T10:00:00Z"]