August 13, 2026
Memory lane or memory scam?
Show HN: MCP Memory – Fast Agent Memory Using Google's OKF and SQLite FTS5
AI gets a memory upgrade, but the comments are asking: isn’t this just fancy search
TLDR: A new tool aims to help AI assistants remember things across chats by saving notes locally and searching them quickly. Commenters weren’t instantly sold: some liked the idea, but many mocked it as little more than ordinary file search with fancier packaging.
A new Hacker News project wants to give AI assistants something they desperately lack: a long-term memory. The idea is simple enough for non-experts—let chatbots save useful notes between sessions, then quickly find them later instead of acting like they woke up with amnesia. This tool stores those memories in readable text files and also keeps a speedy local index so the AI can look things up fast. In theory, that means your coding assistant can remember preferences, project details, and where you left off yesterday.
But the real show started in the comments, where the crowd immediately split into Team Finally, Useful Memory and Team Bro, That’s Just Grep With Extra Steps. One early reply was warm and supportive, with rcarmo basically saying, “Nice, I’ve been building something similar too,” which gave the whole thing a little indie-scene solidarity. Then the skepticism rolled in. Several commenters wanted the same answer in plain English: why not just use normal markdown files? Others pushed harder, asking whether this actually beats existing built-in memory tools “by any metric,” which is Hacker News code for: show receipts immediately.
The funniest and spiciest jab came from the camp that thinks every new AI memory tool is basically the same old idea wearing a fresh outfit. One commenter summed up the mood with a devastatingly dry line about it being another week, another agent memory system that’s basically just search in a folder. Ouch. So yes, the project is neat—but the comment section turned it into a showdown over whether AI memory is finally growing up or just getting better branding.
Key Points
- •MCP-Memory is an MCP server that gives AI agents persistent long-term memory across sessions and chat turns.
- •The system stores memory records in OKF v0.2 Markdown documents with YAML frontmatter and indexes them in local SQLite.
- •SQLite FTS5 is used to support full-text search, key lookups, tag filtering, and content search.
- •The architecture combines human-browseable on-disk `.md` files with hierarchical indexes and a high-performance SQLite search layer.
- •The server exposes four MCP tools: memory_store, memory_retrieve, memory_search, and memory_get_last.