Local-first memory infrastructure for AI agents. AgentMem gives tools like Claude, Codex, Gemini, OpenClaw, Hermes, local coding agents and autonomous scripts a structured way to persist memory between runs.
Instead of losing context every session or scattering state across JSON files, temp folders, prompts, and shell history, AgentMem provides a clean shared memory layer.
Most AI tools today are stateless.
That means they repeatedly need to rediscover:
- what project this is
- where files live
- coding conventions
- current task status
- prior fixes
- known bugs
- open decisions
- team notes
That wastes:
- tokens
- time
- API cost
- focus
- context window space
AgentMem fixes that.
Store useful project context between sessions.
Your memory stays on your machine.
Simple commands that agents and humans can both use.
Organize memory cleanly.
agent/claude/current_task
project/demo/stack
repo/build_command
bugs/auth/login
######################################
######################################
# install globally
cargo install agentmem
# verify install
agentmem --help
# go to your project
cd my-project
# initialize project memory and index codebase
agentmem init
agentmem index
# use re-index after changes to codebase
agentmem reindex
# store memory
agentmem set project/name "My Project"
agentmem set repo/stack "Next.js + TypeScript"
agentmem set project/current_goal "Ship billing v2"
agentmem set agent/claude/current_task "Fix auth bug"
# read memory
agentmem get project/name
agentmem get agent/claude/current_task
# list everything
agentmem list######################################
######################################
cd my-nextjs-app
agentmem init agentmem set repo/stack "Next.js + Supabase" agentmem set project/current_goal "Launch payments" agentmem set agent/claude/current_task "Fix checkout bug" agentmem get project/current_goal