Storing notes is the easy half; the hard half is having the right ones come back. Here recall is shaped by usage — concepts that keep appearing together get more strongly connected, and links you never touch fade out of the way.
A knowledge-graph memory server built on Hebbian learning. Content is saved and queried, and related concepts surface on their own because edges strengthen when two concepts activate together and decay when they do not. It ships with 100+ predefined enterprise concept nodes spanning Systems, Security, Data and Operations.
- Content saved into an associative store and queried back
- Related concepts surfaced automatically, with no manual linking
- Edges that strengthen through co-activation and decay when unused
- 100+ predefined enterprise concept nodes across Systems, Security, Data and Operations
- Stdio transport, so it drops into Claude Desktop, Claude Code or any MCP client
- Three install paths — pip on Windows, pip on Linux and macOS, or Docker Compose for teams
- RAM disk support on Linux through /dev/shm, and a ramdisk profile in Docker Compose
Python with pip for a native install, or Docker Compose for the containerized one; the repository is cloned either way and the server runs on stdio. Behaviour is set through environment variables, or a .env file under Docker. The graph lives where you run it — there is no external service behind it.
Build from source — clone the repository and build it, then point your client at the binary