Hi HN, I’ve been building LLM Memory Notes (LLMMN), a hosted semantic memory layer that gives AI agents persistent, searchable memory via the Model Context Protocol (MCP). You can create “Memories” (containers) and save “Notes” (entries) that are indexed using AI embeddings; agents can then search by meaning (e.g., searching “database error” will find relevant notes about “PostgreSQL timeout” or “MySQL connection failed”) . To get started, sign up on llm‑memory.com, create a Memory in the web UI, and obtain your API token . Add the server to your .mcp.json and use your MCP client’s ReadMcpResourceTool and AddNoteTool to query or add notes ; the docs have a full Quick Start guide . LLMMN is a hosted service (not open source), currently in a public beta with free early access; final pricing is still being determined . We’d love feedback from anyone building agent tooling or experimenting with persistent context layers!