"Nano" matching MCP servers:

Memory for facts that change. A retrieved record is marked SUPERSEDED, with when a later record replaced it, so an assistant does not answer with a value that is no longer true. Single file, numpy the only dependency, runs with a local embedding model or none at all.

Build Nanograph datasources from Markdown or PDF corpora and serve them through MCP or a small HTTP/JSON wrapper.

Hierarchical markdown-based memory system for AI agents. Enables efficient context management by loading only relevant rooms (directories) instead of full memory.

An MCP server that provides a guarded interface to the mem9 persistent memory backend, protecting AI agents against prompt injection, secret leakage, and memory poisoning.