Supermemory — The Memory API That Makes AI Actually Remember You

Your AI forgets everything between conversations. Ask it something you told it last week — blank stare. Supermemory fixes this with a unified memory and context API: it extracts facts from conversations, tracks changes over time, auto-forgets expired info, and gives any AI app a persistent, personalized memory layer. Ranked #1 on all three major memory benchmarks (LongMemEval, LoCoMo, ConvoMem).

Why Not Just RAG?

Traditional RAG retrieves the same documents for all users — it doesn’t know who’s asking. Supermemory is different:

Traditional RAG:
  User asks → Search docs → Same results for everyone → Answer

Supermemory:
  User asks → Search docs + recall user-specific facts → Personalized answer
                              │
                              ├── "User prefers Python over Java"
                              ├── "User's project uses PostgreSQL"
                              └── "User said yesterday: budget is $5K"
                                  (supersedes last month's "$10K")

Key difference: Supermemory tracks facts per user over time, handles contradictions (newer info supersedes older), and auto-forgets temporary context.

Core Capabilities

Quick Start

As MCP Server (Claude Code / Cursor / VS Code)

npx -y install-mcp@latest https://mcp.supermemory.ai/mcp --client claude --oauth=yes

One command — Claude Code gains persistent memory across sessions.

As SDK

npm install supermemory    # or: pip install supermemory

import { SuperMemory } from 'supermemory';
const client = new SuperMemory();

// Store a memory
await client.add("User prefers dark mode and uses vim keybindings");

// Retrieve user profile + search
const result = await client.profile({ user_id: "user123", search: "editor preferences" });

Framework Integrations

Drop-in wrappers for: Vercel AI SDK, LangChain, LangGraph, OpenAI Agents SDK, Mastra, Agno, n8n.

Benchmark Results

The team also open-sourced MemoryBench — a benchmarking framework for comparing memory providers head-to-head.

How It Compares

Architecture: One Ontology, Not Five Systems

Most memory solutions require you to configure separate systems: vector DB for search, graph DB for relationships, key-value store for facts, profile builder for users. Supermemory consolidates everything into a single unified memory ontology:

┌─────────────────────────────────────────┐
│         Supermemory Unified Layer        │
│                                          │
│  ┌──────────┐  ┌──────────┐  ┌────────┐│
│  │ Fact     │  │ Search   │  │Profile ││
│  │Extraction│  │ (hybrid) │  │Builder ││
│  └────┬─────┘  └────┬─────┘  └───┬────┘│
│       │              │            │      │
│       └──────────────┼────────────┘      │
│                      ▼                   │
│           ┌──────────────────┐           │
│           │  Unified Memory  │           │
│           │  Ontology        │           │
│           └──────────────────┘           │
│                      │                   │
│       ┌──────────────┼────────────┐      │
│       ▼              ▼            ▼      │
│  ┌──────────┐  ┌──────────┐  ┌────────┐│
│  │Connectors│  │ Multi-   │  │ Auto-  ││
│  │(Drive,   │  │ modal    │  │ forget ││
│  │ Notion)  │  │(PDF,img) │  │        ││
│  └──────────┘  └──────────┘  └────────┘│
└─────────────────────────────────────────┘

No separate vector DB to configure. No graph DB to maintain. One system.

How LearnAI Team Could Use This

• Persistent course assistants — Give AI tutors memory across sessions so students don’t re-explain goals, misconceptions, or project context.
• Research project continuity — Store evolving facts about datasets, hypotheses, and results so research agents resume accurately.
• Personalized AI literacy coaching — Track each learner’s preferred tools, skill level, and recurring blockers.
• Faculty support workflows — Maintain context across syllabus design, assignment drafting, rubric iteration, and feedback cycles.

Real-World Use Cases

• Personal AI assistants — Remember user preferences, projects, deadlines, and recent decisions across conversations.
• Customer support agents — Retrieve account-specific history without forcing users to repeat context.
• AI coding agents — Preserve repository conventions, architectural decisions, and developer preferences across sessions.
• Enterprise knowledge apps — Combine shared documents with user-specific context for more relevant search.
• Sales workflows — Track changing customer needs, budgets, stakeholders, and follow-up commitments over time.

Links

• GitHub: supermemoryai/supermemory (21K+ stars)
• Docs: supermemory.ai/docs
• Console: console.supermemory.ai
• MemoryBench: Open-source benchmarking for memory providers
• MCP install: npx -y install-mcp@latest https://mcp.supermemory.ai/mcp --client claude --oauth=yes