Verified project record

akitaonrails/ai-memory

ai-memory provides one shared, git-backed markdown long-term memory for 20+ AI coding agent CLIs, including Claude Code, Codex, Cursor, and Gemini CLI. Capture, search, and cross-agent handoffs work with zero LLM calls by default, so no API key is needed.

Project overview

It addresses a concrete pain point in AI-assisted development: agent memory locked to one machine or one tool. By giving more than twenty agent harnesses a single git-backed markdown store with typed claim-once handoffs, it lets work continue across agent and teammate switches without LLM calls by default.

Project type: AI Agent
Use cases: Coding & Development
Deployment: Refer to project documentation
License: MIT

Best for

Developers who use AI coding agents and want a shared, git-backed long-term memory that persists across tools, machines, and teammates without LLM calls by default.

Key capabilities

More than twenty agent harnesses, including Claude Code, Codex, Cursor, and Gemini CLI, feed one shared memory, with typed claim-once handoffs so the next agent continues where the last left off.

Getting started

Install with curl -fsSL https://github.com/akitaonrails/ai-memory/releases/latest/download/install.sh | sh, run ai-memory setup, then ask an agent 'What do you remember about this project?'. Setup difficulty is documented as easy.

Evidence and sources

GitHub project description: Solution for long term memory for agent coding CLIs and to facilitate handoff between different agent vendors
README: More than twenty harnesses (Claude Code, Codex, Cursor, Gemini CLI, OpenCode, Grok, Devin, Kimi, Kiro, and others) feed one shared memory.
Release: v3.4.4
README: The default path uses **zero LLM calls**: capture, search, and handoffs all work with no API key.
README: The source of truth is a git-backed wiki of ordinary `.md` files: `grep` it, open it in Obsidian, edit it by hand, `rsync` it.

Source updated at: 2026-11-08T18:07:44Z

Open original GitHub project