Supermemory
github.com/supermemoryai/supermemory- Category
- AI Tools
- Rank
- No. 107Tools index
- Pricing
- Freemium
- Type
- TOOL
- Builder
- supermemoryai
- GitHub
- 29.4k stars
- Latest release
- server-v0.0.8
- Date
About
A state-of-the-art memory and context engine for AI that automatically learns from conversations, extracts facts, builds user profiles, and delivers persistent memory across AI interactions. It combines RAG, connectors, and file processing into a single system that makes AI assistants remember you.
What it does
Supermemory sits around an AI model call. It retrieves stored user or project context, adds that context to the prompt, then can save the completed exchange. Memories stay separated through container tags, while profiles distinguish durable facts from recent activity.
Why it's ranked here
The project backs its broad pitch with concrete integration code, not only documentation. It supports hosted and local use, several agent frameworks, two major programming languages, and MCP clients. Configurable retrieval modes, storage controls, prompt formatting, and failure behavior make it a serious infrastructure option.
What's good
The wrappers handle both retrieval and persistence around normal model calls. Developers can choose profile, query, or combined retrieval, disable automatic storage, customize prompt formatting, and separate memories by user or project. The Vercel wrapper preserves model properties through a proxy and supports generation and streaming.
Tradeoffs
Framework wrappers require careful identity design because container and conversation identifiers determine memory scope and grouping. Automatic conversation storage is enabled by default in documented wrappers. The Vercel integration uses a fixed five-second retrieval budget, and its default failure policy continues without memories, which can produce inconsistent context between requests.
How to use it well
Use it for assistants that need durable preferences, project history, document retrieval, or recurring user context across sessions. Start with strict container boundaries and explicit storage policy, then test profile and query modes separately. It supplies memory infrastructure, not the underlying language model, agent framework, or application experience.
Technical notes+
The root package.json defines a private Bun and Turbo workspace spanning apps/* and packages/*, with Node 20 or newer required. packages/ai-sdk/tsdown.config.ts emits minified ESM targeting ES2020, plus declarations and source maps. packages/tools/src/vercel/index.ts wraps doGenerate and doStream through a Proxy, applies a fixed 5000 ms retrieval timeout, defaults skipMemoryOnError to true, and saves responses when addMemory is always. packages/tools/src/openai/index.ts requires SUPERMEMORY_API_KEY, containerTag, and customId. Python surfaces are exported from packages/openai-sdk-python/src/supermemory_openai/__init__.py, packages/pipecat-sdk-python/src/supermemory_pipecat/__init__.py, and packages/cartesia-sdk-python/src/supermemory_cartesia/__init__.py.
Observed
- Languages
- The repository contains TypeScript packages and Python SDK packages.
- Package tooling
- The monorepo uses Bun workspaces and Turbo, with Node 20 or newer required.
- Installation surfaces
- The README documents npm, pip, npx local execution, and a standalone local binary installer.
- Interfaces
- Supermemory exposes a hosted API, client libraries, framework wrappers, an MCP server, a consumer app, and local server operation.
- MCP clients
- Documented clients include Claude Desktop, Cursor, Windsurf, VS Code, Claude Code, OpenCode, OpenClaw, and Hermes.
- Framework coverage
- Documented integrations include Vercel AI SDK, LangChain, LangGraph, OpenAI Agents SDK, Mastra, Agno, Claude Memory Tool, and n8n.
- Repository structure
- The root workspace includes application packages and reusable packages under a single private monorepo.
Read from README.md, package.json, packages/ai-sdk/tsdown.config.ts, packages/agent-framework-python/test_real.py, packages/tools/src/index.ts, packages/ai-sdk/src/index.ts, packages/memory-graph/src/index.tsx, packages/tools/src/mastra/index.ts, packages/tools/src/openai/index.ts, packages/tools/src/shared/index.ts, packages/tools/src/vercel/index.ts, packages/tools/src/voltagent/index.ts, packages/openai-sdk-python/src/supermemory_openai/__init__.py, packages/pipecat-sdk-python/src/supermemory_pipecat/__init__.py, packages/cartesia-sdk-python/src/supermemory_cartesia/__init__.py.
What it can do
Extract facts from AI conversations
Conversation data/chat logs → Structured facts and key information
Build user profiles from interactions
User conversation history and behavior data → Comprehensive user profile with preferences and characteristics
Maintain persistent memory across AI sessions
Previous conversation context and user data → Continuous context awareness for future interactions
Process files for AI context
Various file formats (documents, images, etc.) → Processed content ready for AI consumption
Perform retrieval-augmented generation (RAG)
User queries and knowledge base → Contextually relevant responses with retrieved information
Connect external data sources
External APIs and data connectors → Integrated data streams for enhanced AI context
Provide unified context API for AI agents
API requests from AI applications → Complete context stack including memory, profiles, and data
Tags
Tech Stack
Media
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.