Clawlet – AI agent with built-in semantic memory, one binary
Single binary with zero dependencies ships vector semantic memory in SQLite.
local semantic memory for your terminal and coding agents
Local-only agent memory when mem0 and supermemory require servers.
Developers using AI coding agents who want local-first memory
mem0 · supermemory · LangChain memory
I built this because I got tired of telling my coding agent the same things over and over. I also didn’t want to use a hosted memory service like supermemory, or run a local server like mem0. I wanted something small to store notes locally, retrieve them semantically, and keep all the data on my laptop.
It’s not a complex system it’s intentionally simple. You just have a local embedding model that index each memory you add or edit and return the closest embeddings on retrieval (there is a minimum distance threshold so it doesn’t return weak matches).
You can use it like this:
thr add “This repo prefers small PRs with tests“
thr ask “how should i structure this pr?“
I’ve also found myself using it without any coding agent, just as a small notetaking app with semantic retrieval. I might build a tui around it later.
If you try it, I’d love feedback!!!
Single binary with zero dependencies ships vector semantic memory in SQLite.
Cross-project memory for AI agents when single-project solutions already exist.
Single binary with zero deps, but just another agent wrapper atop existing APIs.
Offline semantic memory for agents: SQLite + ONNX embeddings, no API needed.
Hierarchical memory that persists across Claude Code, Cursor, and Windsurf—solve context amnesia.
Vector search and FTS5 in SQLite means no separate vector database server needed.