{ "sym": "Mm", "updated": "2026-08-06", "verdict": "Mem0, for most builders \u2014 the biggest ecosystem (61.7k GitHub stars, 90k+ developers, exclusive memory provider for AWS's Agent SDK), the cheapest real entry ($19/mo), and the fastest shipping cadence (background consolidation 'Dream' landed Aug 4, 2026). Zep wins when you're an enterprise that needs temporal knowledge graphs over business data with provenance, ABAC, and SOC 2/HIPAA. Letta wins when the self-improving agent IS the product, not a feature. Supermemory wins on ingestion volume per dollar and cross-tool personal memory; Cognee when you want to own the whole pipeline open-source. The looming threat is bundling: Claude and ChatGPT now remember natively, Anthropic's developer memory tool is GA, and AWS/Google/Redis sell memory as a metered primitive \u2014 a standalone layer must earn its keep across models and apps, or it's a feature.", "top5": [ { "rank": 1, "name": "Mem0", "maker": "Mem0 (YC S24)", "url": "https://mem0.ai", "docs": "https://docs.mem0.ai", "pricing": "Hobby free (10k adds/mo) \u00b7 Starter $19/mo \u00b7 Pro $249/mo \u00b7 Enterprise custom (on-prem, SSO)", "best_for": "Adding 'remembers the user' to an existing app in an afternoon \u2014 a hosted extraction-and-retrieval API that works with any model and framework.", "why": "The category's center of gravity: 61.7k GitHub stars, 90k+ registered developers, 13M+ package downloads, and API calls that grew 35M (Q1 2025) to 186M (Q3 2025). Raised $24M (seed + Series A led by Basis Set, Oct 2025) and became the exclusive memory provider for AWS's Agent SDK. Still shipping fast: Dream, its background memory-consolidation engine, launched Aug 4, 2026.", "watch": "Its 'SOTA' LoCoMo claims are disputed \u2014 Zep published a rebuttal showing a corrected implementation beat Mem0's setup, and Mem0's own paper had a full-context baseline outperforming it. Retrieval quotas bite: 1k retrievals/mo free and 5k on the $19 tier push retrieval-heavy apps to $249 quickly.", "evidence": [ { "stat": "$24M raised (seed $3.9M + $20M Series A led by Basis Set); API calls 35M Q1 \u2192 186M Q3 2025 (Oct 28, 2025)", "src": "https://techcrunch.com/2025/10/28/mem0-raises-24m-from-yc-peak-xv-and-basis-set-to-build-the-memory-layer-for-ai-apps/" }, { "stat": "61.7k GitHub stars, Apache-2.0; new memory algorithm shipped Apr 2026 (checked Aug 2026)", "src": "https://github.com/mem0ai/mem0" }, { "stat": "Dream background memory consolidation launched Aug 4, 2026; 90k+ developers on platform", "src": "https://mem0.ai/blog" } ], "tile_note": "the default memory api" }, { "rank": 2, "name": "Zep", "maker": "Zep AI", "url": "https://www.getzep.com", "docs": "https://help.getzep.com", "pricing": "Free 10k credits/mo \u00b7 Flex $1,250/yr \u00b7 Flex Plus $3,750/yr \u00b7 Emerging Cos $13k/first year (SOC 2, HIPAA BAA) \u00b7 Enterprise custom", "best_for": "Enterprises whose agents must remember evolving business facts \u2014 temporal knowledge graphs over chat plus CRM/app data, with provenance tracking and access control.", "why": "The engineering-serious enterprise pick: temporal context graphs (open-sourced as Graphiti, 29k stars) with sub-200ms retrieval at 100M graphs, provenance lineage for synthesized facts (Jul 2026), attribute-based access control (Jul 2026), and SSO-gated memory over MCP (Jun 2026). Customers include Zscaler, Samsung, and HoneyBook; S&P Global called it a likely 'de facto partner in this layer of the enterprise agent stack.'", "watch": "No cheap paid tier \u2014 the jump from free to $1,250/yr excludes hobbyists, and the self-hosted Community Edition was deprecated (code moved to legacy/), so the real product is cloud-only. Its 94.7% LoCoMo / 90.2% LongMemEval numbers are vendor-run, like everyone else's.", "evidence": [ { "stat": "94.7% LoCoMo at 155ms, 90.2% LongMemEval at 162ms; 161\u2013168ms retrieval at 10M\u2013100M graphs (vendor, Aug 2026)", "src": "https://www.getzep.com" }, { "stat": "Graphiti OSS: 29k stars, v0.29.2 released Jun 8, 2026", "src": "https://github.com/getzep/graphiti" }, { "stat": "Provenance tracking (Jul 14, 2026), ABAC (Jul 9, 2026), enterprise-SSO memory MCP server (Jun 30, 2026)", "src": "https://blog.getzep.com" } ], "tile_note": "enterprise temporal graph memory" }, { "rank": 3, "name": "Letta", "maker": "Letta (ex-MemGPT, UC Berkeley)", "url": "https://www.letta.com", "docs": "https://docs.letta.com", "pricing": "Free \u00b7 Pro $20/mo (20 stateful agents) \u00b7 Developer $0.10/active agent/mo + $0.00015/sec tool exec \u00b7 Enterprise custom", "best_for": "Building agents whose memory and self-improvement are the product \u2014 stateful agents that learn across sessions, from the researchers who invented the pattern.", "why": "The intellectual origin of the category: MemGPT (Oct 2023) invented LLM virtual context management, and Letta commercialized it with a $10M Felicis-led seed at $70M (Sep 2024). Its March 2026 pivot doubled down on what worked \u2014 Letta Code, a model-agnostic harness with git-backed memory files ('you own the memory, you choose the model'), plus research on sleep-time compute and continual learning that the rest of the field imitates.", "watch": "Strategic churn is real: the Mar 16, 2026 'next phase' deprecated large chunks of the platform (Filesystem, server-side templates, MCP integrations, sleep-time agents, tool rules), and the original server repo is now labeled legacy. Smallest disclosed war chest of the leaders.", "evidence": [ { "stat": "$10M seed led by Felicis at $70M post (Sep 23, 2024); founders created MemGPT at Berkeley's Sky Lab", "src": "https://techcrunch.com/2024/09/23/letta-one-of-uc-berkeleys-most-anticipated-ai-startups-has-just-come-out-of-stealth/" }, { "stat": "Pivot to Letta Code harness with git-backed memory announced Mar 16, 2026; legacy features deprecated by mid-April", "src": "https://www.letta.com/blog/our-next-phase" }, { "stat": "24k stars (Apache-2.0) on the letta repo, now labeled the legacy V1 API server (Aug 2026)", "src": "https://github.com/letta-ai/letta" } ], "tile_note": "memory-first agent harness" }, { "rank": 4, "name": "Supermemory", "maker": "Supermemory", "url": "https://supermemory.ai", "docs": "https://supermemory.ai/docs", "pricing": "Free (~$5 usage) \u00b7 Pro $19/mo \u00b7 Max $100/mo \u00b7 Scale $399/mo \u00b7 Enterprise custom (self-host)", "best_for": "High-volume, cost-sensitive context: memory + RAG + connectors (Slack, Gmail, Drive, GitHub) in one API, plus a personal memory app that follows you across AI tools.", "why": "The fastest riser: 1.5B+ memories stored, sub-300ms recall claims, and a shipping pace that produced a POSIX-compatible semantic filesystem (SMFS, May 28, 2026 \u2014 claimed 55% cheaper agentic retrieval), Context Cloud (May 18, 2026), and default 'dynamic dreaming' consolidation (May 25, 2026) \u2014 all on a $3M pre-seed (Susa Ventures, Oct 6, 2025). Vendor cites internal deployments at Google and Nissan.", "watch": "Tiny funding versus rivals and a split focus (consumer app + infra API). Benchmark leadership claims (LongMemEval, LoCoMo, ConvoMem) and the Google/Nissan logos are vendor-sourced with no independent verification.", "evidence": [ { "stat": "1.5B+ memories saved; sub-300ms recall claimed (vendor, Aug 2026)", "src": "https://supermemory.ai" }, { "stat": "$3M pre-seed led by Susa Ventures, with Browder Capital and SF1 (Oct 6, 2025)", "src": "https://supermemory.ai/blog" }, { "stat": "SMFS semantic filesystem launched May 28, 2026 claiming 55% cheaper agentic retrieval", "src": "https://supermemory.ai/blog" } ], "tile_note": "fast cheap context cloud" }, { "rank": 5, "name": "Cognee", "maker": "Topoteretes", "url": "https://www.cognee.ai", "docs": "https://docs.cognee.ai", "pricing": "OSS free (Apache-2.0) \u00b7 Cloud: Free 1M tokens \u00b7 Standard $2.50/1M tokens + $5/workspace \u00b7 Enterprise custom (BYO cloud)", "best_for": "Teams who want to own the memory pipeline \u2014 an open-source ECL (extract-cognify-load) engine that builds combined knowledge-graph + vector memory over your own databases.", "why": "The credible open-source alternative to hosted memory APIs: 29.8k stars, 5M+ SDK runs monthly, v1 shipped, and real production proof (Bayer runs agentic research memory on it; Knowunity POC'd 40,000 students in 2 days). Deploys self-hosted, Docker, on-prem, or cloud, and plugs into Claude Code, Cursor, LangGraph, and MCP.", "watch": "The smallest commercial operation in the top 5 \u2014 funding undisclosed (Pebblebed and others, amounts unannounced) \u2014 and the ECL pipeline demands data-engineering appetite that Mem0's two-line SDK doesn't.", "evidence": [ { "stat": "29.8k GitHub stars, 5M+ SDK runs monthly, v1 released (vendor, Aug 2026)", "src": "https://www.cognee.ai" }, { "stat": "Bayer production deployment of agentic research memory; Knowunity 40k-student POC in 2 days (vendor case studies)", "src": "https://www.cognee.ai" }, { "stat": "Cloud pricing $2.50/1M tokens + $5/workspace, free 1M-token tier (Aug 2026)", "src": "https://www.cognee.ai" } ], "tile_note": "open-source graph memory" } ], "matrix": { "cols": [ "Approach", "Open source", "Hosted entry price", "Self-host", "Latency claim", "Compliance", "MCP / plugins" ], "rows": [ [ "Mem0", "Extract + consolidate API, opt. graph", "Yes (Apache, 61.7k\u2605)", "Free \u00b7 $19/mo", "Yes (OSS)", "70x vector-search cut (Jul 2026)", "SOC 2 I, HIPAA-ready", "OpenMemory MCP, Claude Code" ], [ "Zep", "Temporal knowledge graph", "Graphiti only (29k\u2605)", "Free \u00b7 $1,250/yr", "No (CE deprecated)", "sub-200ms at 100M graphs", "SOC 2 II, HIPAA BAA", "SSO-gated memory MCP" ], [ "Letta", "Agent-native context mgmt, git-backed files", "Yes (Apache, 24k\u2605)", "Free \u00b7 $20/mo", "Yes", "n/a (in-agent)", "Enterprise SSO tier", "Letta Code harness" ], [ "Supermemory", "Memory + RAG + connectors + SMFS", "Partial", "Free \u00b7 $19/mo", "Enterprise only", "sub-300ms recall", "Enterprise self-host", "Claude/Cursor plugins, MCP" ], [ "Cognee", "ECL pipeline: graph + vector + relational", "Yes (Apache, 29.8k\u2605)", "Free \u00b7 $2.50/1M tok", "Yes (core)", "unpublished", "Enterprise SLAs", "MCP, Claude Code, LangGraph" ], [ "LangMem", "Memory SDK for LangGraph", "Yes (MIT, 1.5k\u2605)", "Free (BYO infra)", "Yes", "n/a", "Via LangGraph Platform", "LangGraph-native" ], [ "Honcho", "Peer modeling + insight reasoning", "Yes (AGPL, 4.9k\u2605)", "Managed api.honcho.dev", "Yes (Docker)", "unpublished", "None published", "API/SDK" ], [ "AgentCore Memory", "Managed events + strategies primitive", "No", "$0.25/1k events", "No", "unpublished", "AWS-grade", "AWS-native" ] ] }, "rules": [ { "if": "You have a working app and just need it to remember users across sessions, this week", "then": "Mem0 \u2014 free for 10k memory-adds a month, $19 after, two-line SDK, works with any model. Watch the retrieval quota, not the add quota." }, { "if": "You're an enterprise whose agents must track facts that change over time \u2014 accounts, policies, patient state \u2014 with audit and access control", "then": "Zep. Temporal graph with provenance and ABAC is the point; budget $1,250/yr minimum, $13k if you need SOC 2 II + HIPAA BAA as a startup." }, { "if": "The agent itself is the product and it must demonstrably learn and improve over weeks", "then": "Letta \u2014 the MemGPT lineage, sleep-time compute, and git-backed memory you can inspect and version. Accept the platform-pivot risk." }, { "if": "You're ingesting everything \u2014 email, Slack, docs, screenshots \u2014 and cost per token retrieved decides the architecture", "then": "Supermemory: connector-heavy, sub-300ms claims, and SMFS cut agentic retrieval costs 55% by the vendor's own math. Verify their benchmarks against your data." }, { "if": "One assistant, one user, one vendor \u2014 'my chatbot should remember me'", "then": "Don't buy a layer. Claude and ChatGPT memory do this natively, and Anthropic's client-side memory tool is GA for developers. A standalone memory layer only earns its keep across models, apps, or agents." } ], "field": [ { "name": "Graphiti", "maker": "Zep (Apache-2.0)", "note": "Temporal knowledge-graph framework under Zep's cloud \u2014 29k stars, v0.29.2 Jun 2026; the OSS on-ramp to Zep", "url": "https://github.com/getzep/graphiti", "oss": true, "entry": "free + your infra", "status": "active" }, { "name": "OpenMemory", "maker": "Mem0", "note": "Local-first MCP memory for coding agents (Cursor, Claude Code, VS Code) \u2014 project-scoped preference recall", "url": "https://mem0.ai/openmemory", "oss": true, "entry": "free", "status": "active" }, { "name": "LangMem", "maker": "LangChain (MIT)", "note": "Memory SDK for LangGraph agents \u2014 only 1.5k stars and increasingly folded into LangGraph's own persistence story", "url": "https://github.com/langchain-ai/langmem", "oss": true, "entry": "free + your infra", "status": "active" }, { "name": "Claude memory + memory tool", "maker": "Anthropic", "note": "Native memory in Claude (Oct 23, 2025, 559 HN points) plus a GA client-side memory tool for developers \u2014 the bundling threat in person", "url": "https://platform.claude.com/docs/en/agents-and-tools/tool-use/memory-tool", "oss": false, "entry": "bundled", "status": "active" }, { "name": "ChatGPT memory", "maker": "OpenAI", "note": "Automatic, continuously-updated memory replaced the manual system; consumer-side only, no developer API exposure", "url": "https://help.openai.com/en/articles/8590148-memory-faq", "oss": false, "entry": "bundled", "status": "active" }, { "name": "AgentCore Memory", "maker": "AWS", "note": "Metered memory primitive in Bedrock AgentCore \u2014 $0.25/1k events, $0.75/1k records/mo stored; commoditization from above", "url": "https://aws.amazon.com/bedrock/agentcore/", "oss": false, "entry": "$0.25/1k events", "status": "active" }, { "name": "Memory Bank", "maker": "Google (Vertex AI Agent Engine)", "note": "Managed long-term user memory for Vertex agents \u2014 Google's answer to the same primitive", "url": "https://docs.cloud.google.com/vertex-ai/generative-ai/docs/agent-engine/memory-bank/overview", "oss": false, "entry": "usage-based", "status": "active" }, { "name": "Redis Agent Memory Server / Iris", "maker": "Redis", "note": "OSS reference implementation (296 stars) graduated into Redis Iris, a managed agent-memory service on Redis Cloud", "url": "https://github.com/redis/agent-memory-server", "oss": true, "entry": "free OSS \u00b7 cloud usage", "status": "active" }, { "name": "MemU", "maker": "NevaMind AI", "note": "14.2k-star lightweight memory-as-markdown-wiki across agents and devices; auto-extracts reusable skills from session logs", "url": "https://github.com/NevaMind-AI/memU", "oss": true, "entry": "free + tokens", "status": "active" }, { "name": "Honcho", "maker": "Plastic Labs", "note": "Peer-modeling memory (AGPL, 4.9k stars) \u2014 background reasoning builds psychological representations of users; managed at api.honcho.dev", "url": "https://github.com/plastic-labs/honcho", "oss": true, "entry": "free + managed tier", "status": "active" }, { "name": "Memobase", "maker": "memodb.io", "note": "Profile-based long-term memory (2.7k stars) \u2014 structured user profiles + time-aware events, sub-100ms retrieval focus", "url": "https://github.com/memodb-io/memobase", "oss": true, "entry": "free + tokens", "status": "active" }, { "name": "MIRIX", "maker": "Mirix AI", "note": "Six-type multi-agent memory (3.5k stars) with screen-observation capture \u2014 personal-assistant angle, local-first", "url": "https://github.com/Mirix-AI/MIRIX", "oss": true, "entry": "free", "status": "active" }, { "name": "Hyperspell", "maker": "Hyperspell", "note": "'Company brain' context graph over 50+ SaaS sources surfaced as an agent-readable filesystem \u2014 overlaps element Kw", "url": "https://hyperspell.com", "oss": false, "entry": "unverified", "status": "active" }, { "name": "Memary", "maker": "community (MIT)", "note": "Knowledge-graph agent memory, 2.6k stars \u2014 last release Oct 2024, momentum gone", "url": "https://github.com/kingjulio8238/Memary", "oss": true, "entry": "free", "status": "fading" }, { "name": "Papr", "maker": "Papr AI", "note": "Former memory-API startup; site now sells AI GTM workflow apps \u2014 memory positioning quietly abandoned", "url": "https://www.papr.ai", "oss": false, "entry": "\u2014", "status": "fading" }, { "name": "MemGPT", "maker": "UC Berkeley \u2192 Letta", "note": "The Oct 2023 paper/project that started the category \u2014 renamed Letta with the Sep 2024 commercialization", "url": "https://github.com/letta-ai/letta", "oss": true, "entry": "\u2014", "status": "renamed" }, { "name": "Zep Community Edition", "maker": "Zep", "note": "The self-hosted OSS memory server that built Zep's following \u2014 deprecated, code moved to legacy/; cloud-only now", "url": "https://github.com/getzep/zep", "oss": true, "entry": "\u2014", "status": "dead" }, { "name": "Motorhead", "maker": "Metal (getmetal)", "note": "Early Rust memory/retrieval server for LLMs \u2014 unsupported since Dec 2023; the category's first grave", "url": "https://github.com/getmetal/motorhead", "oss": true, "entry": "\u2014", "status": "dead" }, { "name": "Rayrift", "maker": "solo developer", "note": "Developer-focused memory layer listed for takeover/acquisition on HN, Jan 31, 2026 \u2014 a marker of how crowded the low end got", "url": "https://rayrift.com", "oss": false, "entry": "\u2014", "status": "dead" } ], "signals": [ { "fact": "Mem0 raised $24M (Oct 28, 2025) on API-call growth from 35M (Q1 2025) to 186M (Q3 2025) and became AWS Agent SDK's exclusive memory provider", "src": "https://techcrunch.com/2025/10/28/mem0-raises-24m-from-yc-peak-xv-and-basis-set-to-build-the-memory-layer-for-ai-apps/" }, { "fact": "Bundling from above: AWS AgentCore Memory meters memory at $0.25/1k events, Google ships Vertex Memory Bank, Redis launched Iris, and Anthropic's developer memory tool went GA \u2014 the primitive is commoditizing", "src": "https://aws.amazon.com/bedrock/agentcore/pricing/" }, { "fact": "Native assistant memory became table stakes: Claude memory launched Oct 23, 2025 (559 HN points); ChatGPT moved to fully automatic, continuously-updated memory", "src": "https://www.anthropic.com/news/memory" }, { "fact": "2026's feature battleground is consolidation: Supermemory made 'dynamic dreaming' default (May 25), Mem0 shipped Dream (Aug 4), Letta published sleep-time compute research \u2014 everyone now sleeps", "src": "https://mem0.ai/blog" }, { "fact": "Benchmark credibility crisis: Zep's rebuttal (May 2025, updated Jun 2026) showed Mem0's LoCoMo comparison used a flawed Zep implementation, and Mem0's own paper had a full-context baseline (~73%) beating its system (~68%)", "src": "https://blog.getzep.com/lies-damn-lies-statistics-is-mem0-really-sota-in-agent-memory/" }, { "fact": "Architecture still unsettled: Letta's Mar 16, 2026 pivot moved memory from server-side databases to git-backed files and deprecated much of its platform API \u2014 the field's founder rethinking the field's premise", "src": "https://www.letta.com/blog/our-next-phase" } ], "notes": "Ranking charter: ecosystem adoption, verified traction, shipping velocity, and pricing accessibility \u2014 explicitly NOT vendor benchmark scores, because every vendor here claims to lead LoCoMo/LongMemEval and the Zep-Mem0 dispute plus LoCoMo's full-context-baseline problem make those numbers unusable for ranking. Conflicts resolved: Supermemory's raise is reported as $2.6M in some coverage and $3M on the vendor blog \u2014 we cite the vendor's own Oct 6, 2025 post ($3M, Susa-led); single-source, flagged. Mem0's site says 62,590 stars while the GitHub page showed 61.7k the same day \u2014 timing/rounding, we cite GitHub. Zep and Cognee funding amounts are undisclosed; treat their runway as unverified. Supermemory's Google/Nissan deployments and benchmark leads are vendor-claimed only. Letta's 24k-star repo is now labeled the legacy V1 server \u2014 star count overstates current-product momentum. Adjacent elements: vector databases (pgvector, Pinecone, Turbopuffer) \u2192 Vd; company wikis and 'company brain' knowledge bases (incl. Hyperspell's overlap) \u2192 Kw; meeting memory (Granola, Otter) \u2192 Mt; MCP servers as distribution \u2192 Mc. Native memory in Claude/ChatGPT is covered here only as the bundling threat, not as picks \u2014 it doesn't cross apps or models, which is this element's whole job.", "sources": [ "https://mem0.ai/pricing", "https://mem0.ai", "https://mem0.ai/blog", "https://github.com/mem0ai/mem0", "https://techcrunch.com/2025/10/28/mem0-raises-24m-from-yc-peak-xv-and-basis-set-to-build-the-memory-layer-for-ai-apps/", "https://www.getzep.com", "https://www.getzep.com/pricing", "https://blog.getzep.com", "https://blog.getzep.com/lies-damn-lies-statistics-is-mem0-really-sota-in-agent-memory/", "https://github.com/getzep/graphiti", "https://www.letta.com", "https://docs.letta.com/pricing", "https://www.letta.com/blog/our-next-phase", "https://techcrunch.com/2024/09/23/letta-one-of-uc-berkeleys-most-anticipated-ai-startups-has-just-come-out-of-stealth/", "https://github.com/letta-ai/letta", "https://supermemory.ai", "https://supermemory.ai/blog", "https://www.cognee.ai", "https://docs.cognee.ai", "https://aws.amazon.com/bedrock/agentcore/pricing/", "https://platform.claude.com/docs/en/agents-and-tools/tool-use/memory-tool", "https://www.anthropic.com/news/memory", "https://help.openai.com/en/articles/8590148-memory-faq", "https://github.com/langchain-ai/langmem", "https://github.com/plastic-labs/honcho", "https://github.com/redis/agent-memory-server" ], "element": { "number": 16, "name": "Memory Layer", "group": "Knowledge & Memory", "essential": false, "edition": "v2026.Q3", "revision": "r7", "license": "CC BY 4.0 \u2014 cite elems.ai", "url": "https://elems.ai/e/mm.html" } }