At a Glance:
Supermemory is a memory and context engine for AI that extracts facts, maintains user profiles, handles contradictions and forgetting, and provides RAG, connectors, and multi-modal file processing through a single API or MCP server.
Overview:
Supermemory is a memory and context layer for AI that gives AI assistants persistent memory across conversations. It automatically extracts facts, builds user profiles, resolves contradictions, and forgets expired information. The project provides two main entry points: end users can install the app, browser extension, or MCP server to give compatible AI clients like Claude Desktop, Cursor, and VS Code persistent memory, while developers can use the API to add memory, RAG, user profiles, and connectors to their own agents and apps. Supermemory ranks #1 on LongMemEval, LoCoMo, and ConvoMem benchmarks.
Key Decision Points:
End-user memory layer vs developer API: Supermemory offers both a no-code app with MCP server and plugins for AI tools, and an API for building memory into custom AI products.
MCP server and client plugins: The open-source MCP server provides
memory,recall, andcontexttools for compatible clients including Claude Desktop, Cursor, Windsurf, and VS Code.Memory is not just RAG: The system tracks facts over time with temporal reasoning and contradiction resolution, while also providing RAG and hybrid search in a single query.
Connectors for external data: Google Drive, Gmail, Notion, OneDrive, GitHub, and a web crawler can auto-sync into the knowledge base with real-time webhooks.
Framework integrations available: Drop-in wrappers exist for Vercel AI SDK, LangChain, LangGraph, OpenAI Agents SDK, Mastra, Agno, and n8n.
Core Features:
Memory extraction and management: Automatically extracts facts from conversations, handles temporal changes, contradictions, and automatic expiration of temporary information.
User profiles: Maintains auto-generated user context combining stable facts and recent activity, retrievable in approximately 50ms via a single API call.
Hybrid search: Combines RAG-based document retrieval and personalized memory context in a single query.
Connectors with auto-sync: Integrates external sources including Google Drive, Gmail, Notion, OneDrive, and GitHub, with real-time webhook-based synchronization.
Multi-modal extractors: Processes PDFs, images with OCR, videos with transcription, and code with AST-aware chunking on upload.
MCP server: Exposes memory save, recall, and context injection tools for compatible AI clients, with open-source implementation.
Use Cases:
Individual users who want their AI assistants to remember preferences, projects, and past discussions across sessions through the app or MCP server.
Developers building AI agents who need to add persistent memory, user profiles, and RAG to their applications through a single API without configuring vector databases or embedding pipelines.
Teams needing scoped memory can organize context by projects (container tags) to separate work and personal information or organize by client or repository.
Open-Source Alternative Value:
Supermemory's MCP server and client plugins are open source, allowing developers to inspect how memory tools are implemented and integrate them into compatible AI clients. The project also provides MemoryBench, an open-source framework for standardized, reproducible benchmarks of memory providers. Developers building on the API get the full context stack — memory extraction, user profiles, RAG, connectors, and file processing — as a managed service that handles temporal reasoning, contradiction resolution, and automatic forgetting without requiring manual configuration of vector databases or chunking strategies.
