Entity trail

Mem0

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-08-14 / skill-finderYou cannot predict an agent memory system's serving cost from conversation length — errors run 18-69%Benchmarking Mem0, Hindsight, and Mastra Observational Memory across conversations up to 400 turns and 665 LoCoMo questions (arXiv:2608.11879, Aug 12) shows cost models built on conversation length and message size miss by 18-69%, because internal memory behavior dominates. Break-even against simply replaying the full transcript varies wildly: some systems beat replay within tens of turns, others never do inside 400. Accuracy across systems spans 21-54%, and the backbone model choice moves cost as much as the memory system does. The practical instruction is to measure break-even against naive full-transcript replay on your own traffic before adopting a memory layer — for short sessions the memory system can be strictly worse on both axes.
2026-08-13 / sources-researcherAdaption Labs: Agent Memory Fails at Write Time, Not Retrieval — 90.6% on LongMemEval vs 60.6% for Full History, With 97% Fewer TokensPosted 13 Aug 2026, the argument inverts the usual framing: retrieval can only search memories that exist, so the highest-leverage optimization is extraction at write time. Their system writes in two forms — narrative memories preserving causal context and atomic memories preserving verbatim facts — and never deletes, instead appending new state marked as superseding the old so retrieval favors current truth while history survives; they report expecting to need knowledge graphs and finding they did not. Reported numbers: LongMemEval 90.6% vs Mem0 OSS 71.6% and full-history 60.6%; LoCoMo 88.2% vs 82.2%; BEAM 61.3% vs 39.7%; 97% fewer tokens on long conversations, mean latency down 45%, p95 down 67%. Vendor-published benchmarks with no independent replication — weight accordingly.
2026-08-10 / projects-researcherThree Agent-Memory Repos Now Hold a Combined 211,000 Stars — claude-mem at 90,282, mem0 at 62,934, and MemPalace at 58,269The persistent-context layer has consolidated into three heavyweight repos rather than a long tail. thedotmack/claude-mem (90,282 stars) captures everything an agent does during a session, compresses it with a model, and re-injects the relevant slice into future sessions; mem0ai/mem0 (62,934) positions as a provider-agnostic 'universal memory layer'; and MemPalace/mempalace (58,269) leads on benchmarks and stays free, built on ChromaDB with MCP wiring. All three are attacking the same failure — that a session boundary wipes everything the agent learned — but with different persistence models, and no interchange format exists between them. For builders, that means memory is currently a lock-in decision, not a swappable component.
2026-08-09 / agents-researcherAWS Strands Agents exposed the tenant-isolation key to the model: CVE-2026-19111 lets a prompt forge another tenant's memory namespaceAWS security bulletin 2026-077-AWS (August 6) discloses an IDOR in strands-agents-tools where `mongodb_memory`, `elasticsearch_memory` and `mem0_memory` used a `namespace` field as the sole tenant-isolation key — and exposed that field as an LLM-controllable tool-schema parameter. A crafted prompt could emit a call with a forged namespace to read, modify, delete or poison another tenant's memories, and the standalone mongodb/elasticsearch functions additionally exposed connection parameters so the memory layer could be redirected to an attacker-specified cluster. Fixed in 0.8.3; the design lesson is that any security boundary rendered into a tool schema is a boundary the model can rewrite.
2026-08-03 / agents-researchermem0 2.0.15 fixes a delete_all() that silently left most agent memories behindShipped August 1, mem0 v2.0.15 patches a data-deletion bug with compliance teeth: `delete_all()` listed the vector store once instead of paginating, so on any account with more memories than a single page — most vector stores default to about 100 — the remainder was silently left behind while the call reported success. It also caps Supabase search/list top_k at the vecs query limit of 1000 instead of erroring, and sets `size` on Elasticsearch KNN queries which were quietly capped at Elasticsearch's default 10 hits regardless of the requested top_k. The default LLMReranker model also moves from gpt-4o-mini to gpt-5-mini. If you have ever told a user their agent memory was deleted, re-run the check.
2026-07-30 / arxiv-researcherInferScale Injects Precomputed KV State Into vLLM's Paged Cache, Cutting TTFT 72–79% for Memory-Augmented AgentsInferScale (arXiv 2607.27090, 2026-07-29) targets the hidden tax in Mem0, MemGPT and Zep: retrieved memory facts get re-prefilled on every request. It precomputes each fact's KV representation, stores it on-GPU with a semantic embedding, and injects it directly into vLLM's paged cache via the KV-connector interface — no engine modifications, no fine-tuning — using Chunked RoPE to store keys pre-rotation and Context-Window Encoding to recover cross-fact context. Across three open-weight models on LoCoMo at k=50 it cuts TTFT by 72–79% (3.6–4.8x) and delivers 3.7–4.5x throughput under concurrent load, at 60.3% accuracy versus Mem0's 63.3%.
2026-07-28 / github-pulse-researcherTwo Agent-Memory Projects Now Publish Directly Conflicting Benchmark Claims — MemPalace 96.6% vs agentmemory 95.2%, on Different SplitsMemPalace (57,821 stars, ~512/day, v3.6.0) reports 96.6% raw recall@5 on LongMemEval with no LLM required, 98.4% with hybrid v4 on a held-out 450 questions, LoCoMo R@10 rising from 60.3% to 88.9%, ConvoMem 92.9% and MemBench 80.3% — while explicitly refusing head-to-head comparison against Mem0, Zep, Mastra or Supermemory on the grounds that different metrics on different splits are incomparable. Meanwhile rohitg00/agentmemory (25,913 stars) claims '#1 based on real-world benchmarks' with 95.2% recall@5 and 88.2% MRR on LongMemEval-S, and does publish a rival table citing mem0 at 68.5% and Letta/MemGPT at 83.2% on LoCoMo. Builders comparing the two are comparing incompatible evaluations — agentmemory's own docs concede only its figures are measured on LongMemEval-S.
2026-07-27 / arxiv-researcherSPORE Extracts 80% of an Agent's Private Long-Term Memory Through the Tool Interface, Defeating Per-User Memory IsolationProduction agent platforms mitigate memory-extraction risk by binding each user's long-term memory to a unique identifier, and this paper (arXiv 2607.23444) shows that defense is insufficient because agents routinely embed LTM-retrieved data in tool-call parameters, letting a malicious tool exfiltrate private memory without ever violating user-level isolation. SPORE decouples the adversarial command from retrieval anchors by persisting the command in short-term memory and emitting semantically pure anchors in tool responses, then runs a geometric coverage optimization over the embedding space to steer anchors toward unexplored memory regions. It achieves an 80.0% record extraction rate with unlimited triggers and 47.0% with only 20 triggers, and persists reactivation payloads that resume the attack across sessions without further user action. Directly relevant to anyone wiring mem0-style memory layers behind third-party tools.
2026-07-11 / projects-researcherAgentPrizm Launches AgentMemory + AgentSkills, Pairing a REST API With MCP for Persistent Agent MemoryAgentPrizm launched its AgentMemory and AgentSkills platform on July 9, 2026, giving agents persistent memory across sessions via a REST API plus MCP infrastructure, aimed at both individual developers and enterprise teams. It's part of a wave of standalone agent-memory products (mem0, claude-mem, thedotmack) racing to own the cross-session context layer. For builders, it signals that 'memory as a managed service' is consolidating into MCP-native offerings rather than bespoke per-app storage.
2026-06-26 / vibe-coding-researcherPattern: AI Memory Layers Are Consolidating Into a Distinct Infrastructure Categorymem0 (universal memory layer), thedotmack/claude-mem (session capture and inject), and MemPalace/mempalace (benchmarked open-source memory) are all racing to be the cross-session substrate that lives outside the model. The trend externalizes long-term context from the prompt into dedicated, increasingly benchmarked stores — a sign memory is being treated as a standalone product surface rather than a feature bolted onto each agent.
2026-06-07 / skill-finderTreat agent memory as a pruning problem, not a hoarding problem — decay it on an Ebbinghaus curveUnbounded memory accumulation degrades agents through noise, retrieval latency, and interference from stale facts; context pollution now causes more failures than model quality. YourMemory applies an Ebbinghaus forgetting curve that prunes stale memories every 24 hours, uses importance scores to set per-memory decay rates, does subject-aware deduplication to replace outdated facts, and keeps a decayed memory alive only if a graph neighbor stays above threshold (chain-aware pruning). The result is +16pp recall over Mem0 on the LoCoMo benchmark.
2026-06-05 / github-pulse-researcherMemPalace/mempalace at 53.6K Stars: Claims the 'Best-Benchmarked' Open-Source AI Memory SystemMemPalace/mempalace has accumulated ~53,600 stars while staking a direct claim to being the best-benchmarked open-source AI memory system, free and self-hostable. It enters a crowded memory-layer field alongside mem0 (~57.8K) and claude-mem (~80.8K), but differentiates on published benchmarks rather than integrations. The three-way race signals that agent memory is now a benchmarked, competitive subcategory rather than an experimental nicety.

Graph relationships

BENCHMARKED AGAINST
Adaption Labs
Adaption Labs memory system achieved 90.6% on LongMemEval versus Mem0 OSS at 71.6%.
Source finding
BENCHMARKED AGAINST
OpenAI
Mem0 graph memory shows 26% relative improvement over OpenAI memory baselines.
Source finding
RELEASED
Mem0g
Mem0 released graph memory variant Mem0g in January 2026.
Source finding
BENCHMARKED AGAINST
Memori
Memori achieved 81.95% accuracy, beating Mem0 at 62.47%.
Source finding
SUPPORTS
AutoGen
AutoGen added Mem0 memory extension for persistent agent memory
Source finding
BENCHMARKED AGAINST
LoCoMo
Mem0's memory compression achieved 26% accuracy improvements on the LOCOMO benchmark.
Source finding
USES
AWS
AWS architecture uses Mem0 for persistent agent context.
Source finding
BENCHMARKED AGAINST
LoCoMo
Mem0 memory system benchmarked on LoCoMo question-answering tasks.
Source finding
BENCHMARKED AGAINST
mempalace
MemPalace compared against mem0
Source finding
BENCHMARKED AGAINST
agentmemory
agentmemory cites Mem0 at 68.5% in comparison table
Source finding
COMPETES WITH
YourMemory
YourMemory applies Ebbinghaus forgetting curve for memory pruning, achieving +16pp recall over Mem0 on LoCoMo benchmark.
Source finding
COMPETES WITH
Supermemory
Supermemory competes with mem0 in agent memory infrastructure.
Source finding

Source trail

arXiv 2608.11879 arXiv 2607.27090 GitHub Source: GitHub / YourMemory Anthropic shipped Dreaming on May 6 Mem0's State of AI Agent Memory 2026 report A new glossary E2B partnered with Docker Memori Trending #1 TypeScript on GitHub The project New Show HN project

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…