II.
AgentProduct overview
Reference · liveagent:supermemory
Supermemory overview
Cloud memory engine and context layer for AI agents. Ranks #1 on LongMemEval, LoCoMo, and ConvoMem benchmarks. Provides automatic memory extraction from conversations, knowledge graph with Update/Extend/Derive relationships, user profile synthesis, hybrid search (RAG + memory), connectors for Google Drive/Gmail/Notion/OneDrive/GitHub, multi-modal processing (PDF, images, video, code), and SMFS filesystem interface. TypeScript + Python SDKs, MCP server, and REST API.
Attributes
displayName
Supermemory
vendor
Supermemory Inc.
homepageUrl
primarySurface
api
supportTier
catalog-backed
productKind
memory-service
stackScope
memory-and-retrieval
description
Cloud memory engine and context layer for AI agents. Ranks #1 on
LongMemEval, LoCoMo, and ConvoMem benchmarks. Provides automatic memory
extraction from conversations, knowledge graph with Update/Extend/Derive
relationships, user profile synthesis, hybrid search (RAG + memory),
connectors for Google Drive/Gmail/Notion/OneDrive/GitHub, multi-modal
processing (PDF, images, video, code), and SMFS filesystem interface.
TypeScript + Python SDKs, MCP server, and REST API.
deepResearchNotes
architectureStyle
cloud-memory-service-with-filesystem-interface
language
TypeScript 64%, MDX 28.9%, Python 6.2%
license
MIT
layerCoverage
L5-agent-runtime
SMFS mount provides transparent memory via filesystem
L12-knowledge-fabric
Knowledge graph, user profiles, semantic search, contradiction resolution
L13-orchestration
Connectors for real-time sync with external services
benchmarks
LongMemEval
#1
LoCoMo
#1
ConvoMem
#1
xAFS
81% accuracy at 10k files, 55% cheaper than baseline
keyComponents
- Memory Engine (fact extraction, contradiction resolution, temporal decay)
- User Profiles (static + dynamic, ~50ms retrieval)
- Hybrid Search (RAG + memory in single queries)
- SMFS (Semantic Memory File System -- NFS/FUSE mount)
- MCP Server 4.0 (Cloudflare Workers + Durable Objects)
- Connectors (7 integrations with real-time webhooks)
- MemoryBench (open-source benchmarking framework)
ourEquivalent
packages/genty/runtime/src/session/memoryExtraction.ts provides
LLM-extracted MemoryEntry objects with category/confidence/tags.
packages/genty/platform/src/session/memoryConsolidation.ts handles
Jaccard-similarity deduplication. packages/genty/platform/src/session/
crossRunState.ts provides key-value cross-run state. Supermemory adds
semantic search, knowledge graph relationships, user profile synthesis,
and multi-modal extraction that we currently lack. SMFS is the primary
integration path -- transparent memory via filesystem operations.
Outgoing edges
competes_with2
- knowledge-fabric-impl:mem0-fabric·KnowledgeFabricImplMem0 as Agent Memory Fabric
- knowledge-fabric-impl:zep-fabric·KnowledgeFabricImplZep as Agent Memory Fabric
has_version1
Incoming edges
part_of2
- knowledge-fabric-impl:supermemory.api·KnowledgeFabricImplSupermemory REST API & MCP Server
- knowledge-fabric-impl:supermemory.smfs·KnowledgeFabricImplSMFS -- Semantic Memory File System