Universal memory layer combining vector search, knowledge graphs, and key-value storage.

from mem0 import Memory
m = Memory()
m.add("I prefer TypeScript", user_id="alice")
results = m.search("preferences", user_id="alice")

OS-inspired virtual memory management for LLMs. Agents maintain core, archival, and recall memory.

from letta import create_client
client = create_client()
agent = client.create_agent(name="my_agent")
# Agent remembers across sessions automatically

Temporal knowledge graph engine where every fact has a validity window.

from graphiti_core import Graphiti
g = Graphiti(uri, user, password)
await g.add_episode(name="pref",
  episode_body="Alice prefers TypeScript")

ECL pipeline combining vector + graph search. Memory in 6 lines of code.

import cognee
await cognee.add("Alice joined in 2023")
await cognee.cognify()
results = await cognee.search("About Alice")

Dynamic memory using Zettelkasten principles — creates interconnected knowledge networks.

# Dynamic Zettelkasten memory
# Auto-links related concepts
# Evolves knowledge graph over time

Claims #1 on LongMemEval, LoCoMo, and ConvoMem benchmarks. Extremely fast and scalable.

# Works with any MCP client
# Claude Desktop, Cursor, Windsurf
# No API key required

Persistent memory for AI coding agents. 25 MCP tools. Fully local, zero cloud.

# 25 tools: capture, search,
# consolidation, timeline,
# auto-surface on session start

Biomimetic 4-network architecture (world facts, experiences, entity summaries, evolving beliefs).

# World facts network
# Experience network
# Entity summary network
# Evolving belief network

OS-inspired memory management for LLMs. Redis Streams scheduling, multi-modal memory.

# MemTensor/MemOS — coding agents
# BAI-LAB/MemoryOS — EMNLP Oral
# agiresearch/MemOS — flexible mgmt

Python library with short/long-term storage, semantic clustering, optional memory decay.

# Short/long-term storage
# Memory decay simulation
# Graph-based associations

LangChain's official memory library. Extracts facts, optimizes prompts, maintains long-term memory via LangGraph's Store.

Role-playing multi-agent framework. Unified Memory class: short-term (ChromaDB), long-term (SQLite), entity, procedural.

The original autonomous agent. Goal-driven with Pinecone-backed vector memory and modular memory components.

Microsoft's framework for agentic AI. Pluggable memory components. Merging with Semantic Kernel into Microsoft Agent Framework.

5000x faster instantiation than LangGraph. Built-in session management, long-term user learning, 100+ integrations.

TypeScript AI agent framework from the Gatsby team. $13M funding. Memory with libSQL and Postgres. 300k+ weekly npm downloads.

Simulates software development roles. Shared memory across PM, architect, engineer, and QA agents.

Open-source AI orchestration with explicit memory control. InMemoryChatMessageStore + Mem0MemoryStore integration.

Visual AI agent builder with 100+ integrations. Drag-and-drop conversational agents with built-in memory.

Enterprise-grade multi-agent orchestration. Agent = LLM + Tools + Memory. Production-ready.

Sub-10ms latency at scale. Built-in embeddings, rerankers. Official MCP server. $130M+ raised.

Sophisticated filtering + vector similarity. Best free tier (1GB forever). HNSW indexing.

GPU-accelerated billion-scale. Multi-vector, hybrid search, sparse+dense. Official MCP plugin.

The developer-friendly default. Local-first, SQLite-backed, rich Python API. Default for many frameworks.

Graph + vector hybrid. Built-in BM25, OpenAI/Cohere integrations. Multi-tenancy native.

Bring vector search to existing Postgres. HNSW + IVFFlat. Zero new infra for Postgres shops.

Serverless, embedded columnar format. Zero server, works inside Python process. Multi-modal native.

Battle-tested billion-scale similarity search from Meta. GPU support. The backbone of many production systems.

The dominant graph database. Native graph storage, Cypher query language, full-text + vector hybrid search. Enterprise support.

Sub-millisecond graph query using sparse adjacency matrices. Redis-compatible. Used by Graphiti (Zep) as primary backend.

High-performance embedded graph DB. In-process, no server needed. Used by kuzu-memory-graph-mcp for sub-3ms recall.

Fully managed graph database on AWS. Supports both property graphs (Gremlin) and RDF (SPARQL). Graphiti-compatible.