Supermemory
SupermemorySupermemory extracts facts from conversations and documents into a graph of memories that it updates, extends and expires as new information arrives, and keeps a profile of each user that an agent can load in one call. Applications reach it through a hosted API, a downloadable local server, an MCP server and plugins for coding agents. The Supermemory team builds it.
Read as the memory engine behind Supermemory's API. The public supermemoryai/supermemory repository holds the MCP server, plugins, framework wrappers and documentation around that engine, not the engine itself.
Openness
1 high confidence- license
- Proprietary (the memory engine and server source are not published)
- source
- closed (the engine runs on Supermemory's cloud or as a downloadable server binary built from a non-public codebase, and no repository publishes it)
- clients
- MIT and Apache-2.0 (the main repository with the MCP server, plugins and framework wrappers is MIT, and the TypeScript and Python SDKs are Apache-2.0, all clients of the engine)
- deployment
- Supermemory cloud by default (a free local server binary limited to 10,000 stored documents, and self-hosting in a customer's own cloud on the Scale and Enterprise plans)
Supermemory's memory engine is not published. It runs on Supermemory's cloud, as a free downloadable server binary built from a private codebase, or in a customer's own infrastructure on paid plans. What the company publishes openly is the code around the engine, such as the MCP server, plugins, framework wrappers and SDKs, which call it without letting anyone build or run it from source.
- https://raw.githubusercontent.com/supermemoryai/supermemory/main/LICENSE recorded 2026-11-20
The public repository's LICENSE is the MIT License text, "Copyright (c) 2025 supermemory".
- https://raw.githubusercontent.com/supermemoryai/supermemory/main/README.md recorded 2026-11-20
The README says the plugins "are implementations of the supermemory API" and that the local server installs as a binary or a prebuilt Docker image: "You don't need to build or clone the server source."
- https://supermemory.ai/docs/self-hosting/local-vs-enterprise recorded 2026-11-20
The local-versus-Enterprise page says of the self-hosted binary "Its server source is not in the public repository" and sells Enterprise as the managed platform with proprietary extraction models.
- https://supermemory.ai/docs/self-hosting/overview recorded 2026-11-20
The self-hosting page says "the downloadable self-hosted server binary is built from a separate, non-public codebase" and "The server binary is not open source"; it is free within a lite license limit of 10,000 stored documents.
- https://supermemory.ai/pricing recorded 2026-11-20
The FAQ answers "Is supermemory open source? Partly." The main repository, MCP server and plugins are MIT and the SDKs Apache 2.0, while "The memory engine is not open source: it runs in our cloud, as the free supermemory local binary, or in your own infrastructure on Scale and Enterprise."
Adoption
3 medium confidenceAdoption is read from monthly downloads of the supermemory SDKs on npm and PyPI, the official clients for Supermemory's API, which reach both the hosted service and the local server. The figure counts installs, automated ones included, rather than active accounts, and leaves out the plugins and the hosted MCP server.
- https://api.npmjs.org/downloads/point/last-month/supermemory recorded 2026-11-20
820,440 downloads of supermemory on npm between 2026-10-20 and 2026-11-18.
- https://pypi.org/pypi/supermemory/json recorded 2026-11-20
The PyPI package is "The official Python library for the Supermemory API", with its repository at supermemoryai/python-sdk and homepage supermemory.ai.
- https://pypistats.org/api/packages/supermemory/recent recorded 2026-11-20
153,138 downloads of supermemory on PyPI in the trailing 30 days.
- https://registry.npmjs.org/supermemory/latest recorded 2026-11-20
The npm package is "The official TypeScript library for the Supermemory API", published from supermemoryai/sdk-ts.
Capability
5 medium confidenceSupermemory performs the whole cycle itself: it extracts facts, links each new one to what it already holds so later facts supersede earlier ones, expires time-bound facts on its own, and serves a user profile beside the fact graph.
- https://raw.githubusercontent.com/supermemoryai/supermemory/main/README.md recorded 2026-11-20
The README says it "extracts facts, builds user profiles, handles knowledge updates and contradictions, forgets expired information", and that temporary facts "expire after the date passes"; memory is scoped by container tags.
- https://supermemory.ai/docs/concepts/graph-memory recorded 2026-11-20
The graph memory page describes "a living knowledge graph of facts" in which memories connect through updates, extends and derives relations, separate from the documents they came from.
Verified 2026-11-20, except some scores, still awaiting re-confirmation