Someone gave Claude Code permanent memory. → 95% fewer tokens per session → Never hits context limits → Picks up exactly where you left off → 46,000 GitHub stars in 48 hours One command: `npx claude-mem install` Completely free. Community-built. Not Anthropic. Claude Code is exceptional. It just kept... forgetting everything. This fixes that.show more

↳

46K stars in 48 hours says a lot about how painful context loss actually is for daily Claude Code users. The "completely free, community-built" angle is also smart — fills a gap Anthropic hasn't officially addressed. Curious how it handles conflicting memories across projects.

@aa22396584 Yeah the demand spike really says everything.

claude-mem is a solid start for session persistence. But if you want actual extended memory architecture, cross-agent, portable, not locked to a particular agent, check out agentmemory: It's designed as a decoupled memory layer that works across harnesses

@MindMatterMoney Stay disciplined even when it feels boring because real growth happens in repetition not in moments of excitement or motivation

@Abid36947870813 @MindMatterMoney That’s where most people give up

@MindMatterMoney Unfortunately

This will save me so much time and money. I currently ask each Claude chat to create a "continuation md" and keep it continually updated throughout and refresh it after every discussion. Takes a lot of my context limit, but I can move to the next chat seamlessly. #claude

Yeah true!

Context limits were honestly the biggest pain

Yeah that alone made it worth trying

Does it work for the full Claude ecosystem?

Follow @charliejhills for more join my free newsletter (49K+ readers)

Big unlock isn’t just fewer tokens—it’s continuity. Once memory survives sessions, the bottleneck shifts to memory hygiene: what gets stored, when it decays, and how you prevent stale context from steering codegen

woah…gonna check this out

You should, it’s actually worth it

i think memory is just the start. Like hermes agent, self-improving ability for ai agent is the future.

super solid work.

Appreciate it

Are you the someone

Imagine how powerful it will be when we use both claude and codex together with no context loss

is there similar stuff for cursor too ? or qwen model ? i wish i can give this to my openclaw too

@charliejhills whoa, if it really keeps context perfectly, i'm gonna give this a shot. context limit struggle is real!

Isso realmente funciona? Estou sendo limitado muito rapidamente, meu plano é 20$

yeah

wait this is actually huge the context limit thing has been killing me on bigger projects

It’s impressive to see how Claude Code is evolving! Excited to see how this will enhance user experience and productivity.

Honestly the forgetting problem was the most frustrating part of longer sessions.

Does this eat up more tokens?

Claude-Mem: filtra fatos e JOGA O RESTO FORA 🗑️ MemPalace: guarda TUDO, porque saber o que falou sem o PORQUÊ não serve pra nada MemPalace hoje é o que melhor resolve o problema real. O resto é filtro "inteligente" que te deixa burro.

wait this is huge does it work across different projects or just within one?

👀

Been running this pattern for a few weeks. The real unlock is not the tokens, it's that your AI never has to relearn your past decisions.

Great..

The biggest question on memory solutions is how do you get rid of real abstractions since not every work really deserves to persist. Aren’t you bloating the filesystem with this? How does it compare to other libs like those based on Karpathy’s knowledge graph?

Wait, this caches your project context between sessions? That's actually clever. The token savings alone would've saved me like 20 bucks this month on my app rewrites.

i found claude-mem still does not avoid claude code to make the same mistake. also it is using a lot of tokens. if you are looking for claude code self-improvement plugin, this might be helpful

If you already love claude-mem give claude-dejavu a try

If that actually works - wow.