supermemory
supermemory gives AI agent developers a persistent memory layer that learns from user, task, or tenant context and injects relevant information into model context in real time. It is available through an API, plugins, and MCP, with deployment options including VPC and air-gapped environments.
One of 671 tools in AI Agents
Who it's for
- AI agent developers who need memory that can carry across models and agent harnesses.
- Teams operating agents in production that need a memory service available through an API, plugins, or MCP.
- Enterprise teams with deployment requirements involving a VPC, data center, or air-gapped environment.
How it fits your workflow
Provide context from users, tasks, or tenants to supermemory. learner-1 extracts and updates learnings, including merging, inferring, and forgetting information in the vector-graph memory database.
When an agent runs, supermemory retrieves prior memory and injects tokens into the agent’s context in real time. The memory layer is designed to work across models and harnesses.
Build through the API, plugins, or MCP. The site describes deployment on a laptop, in a VPC or data center, and in air-gapped environments.
Pricing
Free: A Free plan is named on the page and includes standard PDF extraction; its price and other plan details are not stated.
Free
Not stated
- · Standard PDF extraction
Pro
Not stated
- · Standard PDF extraction
Max
Not stated
- · Standard PDF extraction
Scale
Not stated
- · Advanced PDF extraction with OCR and descriptions of figures and diagrams
Enterprise
Not stated
- · Advanced PDF extraction with OCR and descriptions of figures and diagrams
Prices checked on Oct 19, 2026 from the vendor's site. They can change; confirm before you buy.
Key features
- Continual learning with learner-1
- learner-1 extracts and updates information from the context of users, tasks, and tenants, and can merge, infer, and forget information in the memory database.
- Vector-graph memory database
- Memory is stored in a vector-graph database. The site describes learnings as interconnected and linked to entities, sources, and moments.
- Real-time context injection
- Relevant prior memory is brought back to an agent by injecting tokens into its context in real time.
- Model and harness interoperability
- The memory layer lives outside model providers and is designed to work across models and agent harnesses.
- Multiple access methods
- The product is available through an API, plugins, and MCP.
- Deployment options
- The site describes running supermemory in a data center, a VPC, on a laptop, or in an air-gapped environment.
Works with
- MCP
- API
- Plugins
Limitations to know
- Free, Pro, and Max plans use standard PDF extraction; advanced extraction with OCR and figure or diagram descriptions is listed for Scale and Enterprise.
- The supplied page does not state plan prices.
Alternatives to consider
-
Mem0
Choose it if you want a persistent memory layer with an SDK and API for adding, updating, and retrieving memories across user interactions and sessions.
-
Recalld
Choose it if you want a memory layer specifically described as storing conversations and documents, reconciling new facts, and offering a native MCP server.
-
DocuMemory
Choose it if your main need is turning AI chat sessions into searchable engineering memories for later recall, with hosted and self-hosted options.
Getting started
- Open the site’s “Build with supermemory” entry point.
- Choose API, plugins, or MCP as the access method.
- Use the provided setup prompt, which can be opened in ChatGPT, Claude, Cursor, or Grok.
- Consult the self-hosting documentation or talk to sales about deployment requirements.