Mem0
Universal memory layer for AI agents. Three concurrent stores (vector + graph + KV); LLM-extracted facts; concurrent retrieval via ThreadPoolExecutor.
At a glance
- Type
- Vector + graph + KV (hybrid)
- Tier
- T1
- Section
- Dedicated memory layers
- Created
- 2023-06
- Latest release
- openclaw-v1.0.11 2026-04-29
- License
- Apache-2.0
- GitHub
- 54.9k★ +1.6k/mo Python
- Pricing
- Free + paid
- Funding
- $24M total $150M val Series A · 2025-10
Taxonomy
- storage
- vector
- retrieval
- similarity
- persistence
- long-term
- update
- extraction
- unit
- fact
- governance
- opaque
- conflict
- llm-arbitrate
When to use
Optimised for: developer experience + universal memory layer (model-agnostic, multi-store)
Anti-fit: no anti-fit explicitly stated
Pros & cons
Pros
Hybrid (vector + graph + KV) gives the most architectural flexibility of any memory layer; AWS Agent SDK exclusivity and 51k★ make it the field's de-facto reference.
Cons
LOCOMO benchmark numbers were publicly disputed by Zep in counter-analysis; LLM-extraction approach risks dropping facts that don't fit the prompt.
Claims & capabilities
~51k★. $24M Series A (Oct 2025) at $150M. Exclusive memory provider for AWS Agent SDK. Reports 26% improvement over OpenAI on LOCOMO; 91% lower p95 latency vs full-context.
Technical surface
- API surface
- REST, SDK: Python, Node.js
- Backend storage
- hybrid (vector + graph + KV)
- Deployment
- Both
- Embedding model
- multiple supported
- Multi-tenancy
- Logical namespace per (user_id, agent_id, run_id); self-hosted/on-prem deployment available for tenant isolation
- MCP
- native (first-party) — official mem0-mcp server
- A2A
- not documented publicly
- OpenTelemetry
- via adapter — AgentOps integration
Compare Mem0 with…
Similar systems
Other dedicated memory layers in the catalog, ranked by inbound references.
- Zep & Graphiti T1
Bi-temporal knowledge graph (event time + ingestion time). Strong on chronological reasoning and contradiction tracking. Graphiti is the open-source core.
- Cognee T1
"Extract–Cognify–Load" pipeline that turns raw input into a typed, queryable knowledge graph for agent recall.
- Hindsight (Vectorize) T1
Standalone memory service from Vectorize. Open source. Biomimetic four-network design (World, Bank, Observation, Opinion). Ships an MCP memory server.
- Memvid T2
Single-file memory layer (one .mv2 file). No DB, no server. Append-only sequence of immutable Smart Frames with timestamps + checksums. Native Rust core (rewritten from Python).
- Supermemory T1
Memory engine with API, app, browser extension, and MCP server. Extracts facts, tracks updates, resolves contradictions, auto-forgets expired info. Plugins for Claude Code, OpenCode, OpenClaw, Hermes.
- AI Singapore SEA-LION T2
SEA-LION-Embedding (March 2026): retrieval + reranking models contrastively trained on 245M text pairs across 10 SE Asian languages. SEA-BED benchmark (169 datasets). SEA-LION v4 (Gemma-based) at 128K context with native function calling.
Related systems
References (4)
- Amazon Neptune Analytics builds on — AWS architecture: build persistent memory with Mem0 open source, Amazon ElastiCache for Valkey and Amazon Neptune Analytics
- Amazon Neptune Analytics integrates with — Neptune Analytics Mem0 integration GA 2025
- Neo4j depends on at runtime — adjacent-infrastructure cell: BYO LLM provider; bundles vector store (Qdrant default) and graph store (Neo4j optional)
- Qdrant depends on at runtime — adjacent-infrastructure cell: BYO LLM provider; bundles vector store (Qdrant default) and graph store (Neo4j optional)
Referenced by (19)
- AgentOps integrates with — When Mem0 is connected, gains Memory Operation Timeline, Search Analytics, Memory Growth tracking, Error Tracking per memory call.
- Amazon Neptune Analytics integrates with — Vector index on graph nodes queryable via openCypher. Mem0 integration GA 2025
- Amazon Neptune Analytics depends on at runtime — aph nodes queryable via openCypher. Mem0 integration GA 2025; Cognee integration for agentic RAG. Combines semantic
- AutoGen Memory integrates with — ListMemory chronological context + teachable agents that vectorise corrections. Integrates with Mem0/Zep rather than building deep memory natively.
- Bedrock AgentCore (AWS) depends on at runtime — adjacent-infrastructure cell: AWS Bedrock; Mem0 (as memory provider); Strands Agents
- CrewAI depends on at runtime — e. Memory available via Mem0 / Letta adapters (CrewAI Memory cross-listed).
- CrewAI Enterprise depends on at runtime — adjacent-infrastructure cell: OSS CrewAI (already in catalog); typically paired with Mem0 / Zep for memory + Langfuse / LangSmith for observability
- CrewAI Memory integrates with — Memory subsystem inside the CrewAI orchestration framework; integrates with Mem0 for the long-term tier.
- CrewAI Memory depends on at runtime — adjacent-infrastructure cell: requires CrewAI; integrates Mem0 for long-term
- DSPy History integrates with — Official Mem0+DSPy tutorial.
- FalkorDB builds on — G SDK. Graph-memory backend for Mem0; MCP server via Graphiti integration; context graphs for lo
- FalkorDB depends on at runtime — G SDK. Graph-memory backend for Mem0; MCP server via Graphiti integration; context graphs for lo
- Haystack Memory (deepset) integrates with — 18k+ stars; production via deepset Cloud; integrates Mem0 as a memory store alternative.
- Langflow Memory integrates with — Bundled Mem0 + Redis integrations for external memory layers.
- Mem0 MCP (official) builds on — Mem0's managed cloud API exposed as MCP tools.
- OpenMemory MCP builds on — Local-first self-hosted Mem0 variant. Shared memory layer bridges Claude Desktop, Cursor, and Windsurf in one session.
- Ratine integrates with — Free / open source. Auto-detects Mem0 local, AutoGen, custom layouts.
- Strands Agents Memory (AWS) integrates with — Memory via Mem0 integration (ElastiCache for Valkey + Neptune Analytics) or Bedrock AgentCore Memory. AWS+Mem0 partnership.
- Strands Agents Memory (AWS) depends on at runtime — ni, OpenAI). Memory via Mem0 integration (ElastiCache for Valkey + Neptune Analytics) or Bedrock Age