GitHub radar
Agent Memory: Persistent Memory for Claude Code
A local first, agent agnostic, long term memory runtime for Claude Code and Codex CLI. Memories saved as plaintext Markdown files on your computer. No API keys. No cloud services.
Agent Memory is a local-first, agent agnostic, long term memory runtime for AI agents. Memories are saved as plaintext Markdown files on your computer. Greppable, git trackable, portable. There’s also an SQLite index for a rebuildable cache. It’s safe to delete. Uses conversation state to write memories, not agent timing. Use a Manage pass outside of conversations to age/archive memories based on value, can only suggest deletions (never perform them). 52.9% LongMemEval-S accuracy (vs. 35.8% MemCore, 5.8% no memory). Linux, Windows, macOS (Apple Silicon) support.
Why a vibe-coder should care
agent-memory saves you from having to re-explain context to Claude Code or Codex when working on the same projects day to day. Your agent will remember your technology stack, decision making, preferences. No subscriptions, no cloud accounts.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up agent-memory from https://github.com/tigerless-labs/agent-memory — follow the README, install via uv, wire it into Claude Code with `--host claude-code`, initialise the store, and show me it works by writing a memory and recalling it.
Runs on a regular laptop — Python 3.12 and uv required.
Open on GitHub▌ More finds