I Built a Workaround for Claude's Memory Problem. Then I Retired It.
- 13 Mar, 2026
I use Claude Code for everything. Coding, filing receipts, managing my productivity system. But every session started from scratch. Zero memory of what happened yesterday, last week, in a completely different project.
The only way to reference past work was to open another Claude Code window, switch to that project folder, and ask Claude about it there. Context in one silo. Knowledge in another.
Then I installed claude-mem.
A second brain across every project
claude-mem is a plugin that silently records observations after every tool call. Not raw output — compressed, categorized notes. Bug fixes, decisions, discoveries. All stored locally in a SQLite database on your machine.
Here’s the part that got me. The memory works across projects. I can be knee-deep in frontend work and ask “what did we change on the API last week?” It just knows. No folder switching. No second window. One brain.
If you’ve ever used a RAG system to make documents searchable, this is the same idea applied to your own work. Except you don’t have to set anything up. The plugin handles the indexing automatically.
The context window problem
Anyone who uses Claude Code for long sessions knows the pain. Context gets tight. The built-in “compact” feature compresses your conversation, but you don’t control what survives.
So I built my own workaround. A custom skill that captured what I wanted explicitly — key decisions, files touched, gotchas — and copied it to my clipboard. I’d clear the session, paste that summary into a fresh one, and keep going.
It worked. But it was duct tape.
A single summary flattened everything — which files we touched, what decisions we made, what gotchas we hit — into a paragraph that lost most of the nuance.
Progressive disclosure beats brute force
claude-mem takes a fundamentally different approach. Instead of one big summary at the end, it captures everything incrementally, in real-time, while context is plentiful.
The retrieval follows a three-layer pattern:
- Search — returns a lightweight index of observations. Just IDs, timestamps, types, and titles. Maybe 50-100 tokens per result.
- Timeline — shows context around interesting results. What happened before and after a specific observation.
- Fetch — pulls back full details only for the specific observations you actually need.
This is progressive disclosure applied to AI memory. You don’t pay for what you don’t need. The index is cheap. The details are on-demand.
When you clear context and start fresh, claude-mem re-injects that lightweight index of everything that happened. You pull back only what’s relevant. No wasted tokens on context you’ll never reference again.
I retired my workaround. And my token usage went down.
Why this matters beyond convenience
There’s a broader point here. I wrote recently about how companies are starting to measure AI token consumption as a proxy for productivity — repeating the same mistake we’ve made with every vanity metric before it.
claude-mem is interesting because it works in the opposite direction. It makes you more effective while consuming fewer tokens. The memory means Claude doesn’t have to re-discover things it already learned. The progressive disclosure means you’re not stuffing your context window with information “just in case.”
It’s the difference between carrying every book you own in a backpack versus having a library card. Both give you access to knowledge. One breaks your back.
Getting started
claude-mem is a Claude Code plugin by Alex Newman. Install it, and it starts working immediately. No configuration required for the basics.
Everything is stored locally. Your observations, your decisions, your context — all on your machine. Nothing leaves unless you tell it to.
If you use Claude Code for anything beyond one-off questions, this is worth trying. Especially if you work across multiple projects and find yourself wishing Claude remembered what you did yesterday.
It does now.
Stay Ahead in Product Management!
Ready to elevate your product management game? Join our community of passionate professionals and be the first to receive exclusive insights, tips, and strategies directly in your inbox. Subscribe now to our newsletter and ensure you're always in the loop with the latest product management trends and wisdom bombs!
