Supermemory

Supermemory is an open-source and self-hosted web app and server that provides a memory and context layer for AI applications. It allows AI assistants to retain information across different conversations by extracting facts, building user profiles, and managing knowledge updates. The software is available as a web application, a browser extension, and a server that can be deployed locally or via a single binary for zero configuration setups.

Users can integrate the system into existing AI workflows using an MCP server or a dedicated API. It supports various clients including Claude Desktop, Cursor, Windsurf, and VS Code. The system manages memory through project-based container tags, which allows users to separate different contexts such as work, personal data, or specific client repositories to prevent context leakage.

Key features

• Automatic extraction of facts and preferences from conversations
• Hybrid search combining RAG and personalized memory in one query
• Auto-maintained user profiles with static facts and dynamic activity
• Connectors for Google Drive, Gmail, Notion, OneDrive, and GitHub
• Multi-modal extractors for PDFs, images, videos, and code
• MCP server for integration with AI code editors and assistants
• Support for offline operation using Ollama
• Project-based scoping via container tags

The architecture is built with TypeScript, Tailwind CSS, and Postgres, utilizing Drizzle ORM and Cloudflare Workers. It provides a single API for memory, RAG, and file processing, removing the need for manual vector database configuration, embedding pipelines, or custom chunking strategies. The system includes drop-in wrappers for frameworks such as the Vercel AI SDK, LangGraph, and n8n, enabling developers to inject user profiles and relevant memories into system prompts with low latency.

Supermemory is a comprehensive context stack for developers building AI agents and individuals seeking a persistent personal knowledge base.