Supermemory
Context engineering platform providing the memory layer for AI agents with retrieval, user profiles, and multi-source data connectors.
Last verified:
What is Supermemory?
Supermemory is a developer-focused Memory API and platform that provides long-term and short-term memory infrastructure for AI agents, enabling applications to store, retrieve, and reason over user-specific context and events. It converts unstructured interactions (chats, files, emails, links) into structured memories, builds a living knowledge graph, and returns relevant context to LLMs to produce personalized, consistent responses. Key features include automatic content extraction, semantic search and retrieval, connectors for apps and data sources, a Model Context Protocol (MCP) server for persistent cross-application memory, and an optional self-hostable single-binary deployment for on-prem or private use. Supermemory targets AI product developers, researcher teams, and companies building chatbots, personal assistants, and any LLM-powered app that needs persistent, evolving user memory to improve personalization, retain context across sessions, and enable managed RAG workflows.
Supermemory pricing
Pricing model: Freemium
The website and docs point users to the Supermemory Developer Console for account creation and API key issuance and describe a free developer tier and paid plans available through the console; the docs highlight that you can sign up at the console to see details, obtain API keys, and start with a free tier before upgrading. Self-hosting is supported via a single-binary install command for local/private deployments, while higher-volume or production use is covered by paid plans managed through the account console. Exact plan names, quotas, and per-feature inclusions (connectors, MCP server, managed RAG limits) are surfaced in the console billing pages rather than fully enumerated on the public documentation pages.
Supermemory pros
- Provides both short-term and long-term memory storage specialized for AI agents
- Automatic extraction of memories from conversations and uploaded content
- Builds and maintains user profiles combining static facts and dynamic context
- Creates a knowledge graph (living memory) rather than simple document storage
- Semantic search and relevance thresholds to return contextually relevant memories
- Managed RAG capabilities to integrate retrieved memory into LLM prompts
- Connectors for common apps and file sources to sync external data
- Model Context Protocol (MCP) server to share memory across applications
- Self-hostable single-binary option for private or on-prem deployments
- Developer-friendly SDKs and quickstart examples for Python and JavaScript
- Memory router that lets you add unlimited memory with a URL change to existing LLM providers
- Fine-grained control via API keys and programmatic ingestion
- Optimized for production: scalable, performant, and affordable infrastructure
- Supports multiple content types (files, emails, chats, PDFs, links) for multimodal memory
- Tools and docs to create thresholds and filters for relevance scoring
- Open tooling and documentation and an active GitHub repo for integration examples
Supermemory cons
- Website docs indicate developer focus; not a turnkey end-user consumer app
- Requires integration work and API key management to use with existing LLM providers
- Some advanced features (MCP, self-hosting) add operational complexity to deploy and manage
- Pricing model details are described across console/docs; may require signup to view exact plan limits
- RAG and connector setups require configuring provider API keys and additional infrastructure
- On-prem/self-hosted binary still requires ops expertise and security review
- Not primarily a non-technical no-code product—some engineering is needed to extract full value
- Feature set and roadmap imply active development; breaking changes or rapid iterations may require frequent updates
Frequently asked questions about Supermemory
How do I get started and obtain an API key?
Sign up at the Supermemory Developer Console, navigate to API Keys and create an API key, then install the SDK (pip/npm) and set the SUPERMEMORY_API_KEY environment variable to begin adding and retrieving memories.
What kinds of data can Supermemory ingest as memories?
Supermemory can ingest unstructured content such as chat messages, documents, emails, PDFs, links, and application data streams and it extracts memories and insights from these sources to populate the knowledge graph.
Can I run Supermemory locally or on-premises?
Yes — Supermemory provides a self-hostable single-binary installation option so teams can run the service locally or in private infrastructure for data residency or security requirements.
How does Supermemory return relevant context to my LLM?
Supermemory performs semantic search over stored memories and returns relevant context according to configurable relevance thresholds and retrieval parameters so your LLM prompt receives focused, personalized information.
What is the Model Context Protocol (MCP) and why use it?
MCP is a protocol Supermemory supports via an MCP Server component that enables persistent memory sharing across different applications and LLM clients so that memories captured in one app become available to others using the MCP-compatible layer.
Which LLM providers work with Supermemory?
Supermemory is designed to sit in front of major LLM providers — docs and examples show usage with OpenAI, Anthropic, Google Gemini, and Groq by updating base URLs and including both provider and Supermemory API keys in requests.
How does Supermemory handle relevance filtering and thresholds?
When querying, you can use the threshold parameter and other retrieval settings to filter search results by relevance score so only sufficiently relevant memories are returned to the model for prompt assembly.
Is Supermemory suitable for production-scale applications?
Yes — the documentation describes the platform as scalable, production-ready and performant, with managed RAG features and an API-first offering intended for production deployments.
Does Supermemory provide SDKs and code examples?
Yes — Supermemory provides SDKs, quickstart guides, and code snippets for Python and JavaScript, along with step-by-step instructions for creating API keys, installing the SDK, and making first API calls.
How do connectors and syncing work for external data sources?
Supermemory includes connectors to sync external apps and storage (like cloud drives and productivity apps) so external documents and data streams can be ingested and converted into memories for semantic retrieval.