🧠 agent-memory

A small self-hosted, mem0/Supermemory-style long-term memory layer for AI agents — extracts atomic facts from text, embeds them, and retrieves them ranked by relevance, importance and recency. Built on Cloudflare Vectorize + Workers AI + D1 + Claude Haiku 4.5.
Source on GitHub →

Teach it something POST /remember

Paste a few sentences — it extracts the atomic, memorable facts (not the whole blob) and stores them.


Ask it something POST /recall

Query in plain language — results are ranked by semantic similarity × importance × recency, not similarity alone.