Claude-Mem Feedback

[thedotmack]

Hello Claude-Mem...bers? 😂

Thank you so much for trying Claude-Mem! Creating this has been a labor of love for me, and I am so grateful for such a positive response from the community.

I am posting this discussion to serve as a place for anyone to post feedback, comments, concerns, hate mail, love letters, anything you'd like to share about how Claude-Mem is working.

Please feel free to write positive or negative feedback! It is all valuable and will always be considered when it comes to improving Claude-Mem for all of us. 😃

Thanks,
@thedotmack

[felores]

Hi, I think this is a great idea overall, is in the development roadmap a way to tweak the system memories per project? I find counterproductive to have it pop in my second brain root, because theres where I manage all my projects and every new chat gets context pollution, but is not only in my second brain where it gets annoying, I also work in different parts of repositories at the same time, for example working on frontend work and backend work in other instance but as you can imagine each thread start getting polluted by the other tasks happening in the other areas

I could envision a way to enable memories per feature, and the ability to also turn on and off memories per repository, that would make it really usable for me.

Cheers!

[thedotmack]

Hi @felores – the memories are organized by project, it uses the folder root you start working in as the project name, are you saying you have one Claude orchestrating multiple projects, but all the memories are showing up in your orchestration agent history?

I'm not sure what you mean by "pop in my second brain root"

[thedotmack]

Is your issue that when you work at the parent /SecondBrain/ level, then later start Claude in /SecondBrain/ProjectA/, the subfolder project doesn't see the relevant memories from the parent-level work?

If so, I think we may be able to filter parent project memories by file path to inject only the relevant ones into the subfolder's context.

[thedotmack]

And being able to customize the context injection based on the project you're working on.... That's a better workflow right? So workflow could be like:

Set claude-mem context to ProjectA

skill running...

Done! Context injection after /clear will be from ProjectA

---

work work work
memories memories memories

---

Great job Claude! Let's work on ProjectB next session

skill running...

Done! Context injection after /clear will be from ProjectB

[felores]

I think the main issue for me is that I cant deactivate it on a per project basis, correct me if Im wrong, I use claude code for software development but also as a second brain and not being able to disable in the second brain is a problem because it injects unrelated memories to every new chat, unless Im using it wrong

[thedotmack]

To keep things project scoped, it must start from the root of the project. Then it scopes it automatically.

That said, it seems like you work from a "second brain" folder that has subfolders with projects? And you'll prompt this second brain claude, it will work on one project, then maybe another... and now you have mixed memories?

This can be achieved with filtering, it just needs to be coded in.

What does your project folder structure look like?

[felores]

root > projects > category > [project_root]