The Problem: Claude Forgets Everything

If you've been using Claude Code for any real project work, you've probably noticed the limitation: every new conversation starts from scratch. Claude doesn't remember what you worked on yesterday, the decisions you made last week, or the architecture you spent an hour explaining.

That changes with claude-mem, a community plugin that gives Claude Code persistent memory across sessions. It stores observations, plans, and context in a local database so future conversations can pick up right where you left off.

Here's how to set it up in about 60 seconds.

The 4-Step Setup

Step 1: Add the Plugin Marketplace

First, register the marketplace that hosts claude-mem. Open Claude Code and run:

/plugin marketplace add thedotmack/claude-mem

You should see: Successfully added marketplace: thedotmack

Step 2: Install the Plugin

Now install claude-mem from that marketplace:

/plugin install claude-mem

You should see: Installed claude-mem. Run /reload-plugins to activate.

Step 3: Reload Your Plugins

Activate the plugin by reloading:

/reload-plugins

This loads claude-mem's tools and slash commands into your session. You'll see a confirmation with the number of plugins, commands, and hooks loaded.

Step 4: Start Using It

That's it. You now have access to these new commands:

Try it out by asking Claude to remember something, or run /make-plan on your next feature and watch how it breaks the work into phased steps.

Why This Matters

Memory changes the dynamic with Claude Code completely. Instead of re-explaining your project every session, Claude can recall past decisions, reference earlier work, and build on context that already exists. It's the difference between working with a new contractor every day versus a teammate who's been with you the whole time.

The plugin is open source and the memory stays local on your machine. No data leaves your environment.

Get the Plugin claude-mem

Browse the source, report issues, or contribute on GitHub.

View on GitHub

What to Try First

Once you've installed it, here are a few things to experiment with:

  1. Plan a feature: Run /make-plan and describe something you want to build. Watch how it breaks the work into phases with clear documentation.
  2. Execute the plan: Run /do and let subagents carry out each phase while tracking progress.
  3. Search your history: After a few sessions, try /mem-search to pull up past context. It's surprisingly useful when you come back to a project after a break.
  4. Explore your codebase: Use /smart-explore to quickly understand code structure without reading entire files.

The more you use it, the more useful the memory becomes. Each session adds to the knowledge base that future sessions can draw from.

Happy building.