A couple of weeks ago: an idea.
Remember the feeling to the AI,
"i have already told you this!"
Today: Memra is live. πͺ
https://t.co/RHrxY4Exdl
persistent memory for claude code, cursor, and every ai tool in your stack. local-first. cloud-synced. eu-hosted.
Solo. Bootstrap. No VC.
Brutal feedback welcome.
(Be gentle π)
#buildinpublic #indiehackers
Yeah static markdown was my v1 too, worked great until I started bouncing between devices.
Laptop on a train
desktop at home
phone for quick fixes...
My context kept fragmenting and I'd end up with 3 different versions of "the truth."
That's actually why I built it as an MCP server with a hosted backend.
Now my agent pulls the same memory whether I'm on the road, on a plane, or borrowing a friend's machine.
Same project context, same decisions, same conventions, everywhere.
Files are great when you're at one desk.
Once you move around, you need it to follow you.
Feel free to use the app, and please let me know if you need any support. I hope you can see the benefits this bring to AI Coders. Try it for free π
Look here, i even forgot that i did not work on the 2026-04-10, and i did not know that, and i even had a typo on the date, but Claude fully understand and could get and check for memories that has been saved.
@pmitu https://t.co/DzRKnEEyYL
I got tired of re-explaining my project to Claude Code every session, so I built a memory MCP server for it, a persistent memory!
It stores decisions, patterns, and conventions.
Your AI remembers your project context across every session.
Feedback welcome!
Gave Claude Code persistent memory today.
Took 30 seconds.
claude mcp add memra
Claude suddenly has 7 new tools:
remember
recall
search
history
supersede
list
get
Started storing decisions automatically.
Recalled them on next session.
MCP is powerful
https://t.co/DzRKnEEyYL
A couple of weeks ago: an idea.
Today: Memra is live. πͺ
https://t.co/nU264TwocJ
Solo. Bootstrap. No VC.
Brutal feedback welcome. (Be gentle π)
#buildinpublic#indiehackers
@charliejhills That's actually why I ended up building Memra (https://t.co/DzRKnEEyYL) its a memory layer any MCP-aware tool can hit, so Claude Code, Cursor, Droid and Claude Desktop all share one brain.
Your Obsidian vault stays the human-readable layer, Memra handles the retrieval underneath
@RoundtableSpace I got tired of re-explaining my project to Claude Code every session, so I built a memory MCP server for it, a persistent memory!
It stores decisions, patterns, and conventions.
Your AI remembers your project context across every session.
https://t.co/nU264TwocJ
Built Memra to fix exactly this.
Memory layer for LLMs.
Plug it in via MCP, agents remember across sessions and across tools.
Free tier EU-hosted:
https://t.co/DzRKnEEyYL
I hope this is useful for some of you. If not, tell me why.
#mcp#claude
My coding agent just forgot everything. Again.
The API spec i walked it through yesterday.
The three refactors i already rejected.
The name of my actual database.
Gone.
This is the #1 thing slowing down AI-assisted dev right now!
#aicoding#vibecoding
@CestIvan@victor_bigfield yes, weeks, months, as long as you have the memory.
I just asked this question:
"look inside memra, what did we do on 2006-04-10?"
empty context window!
wrong year as well, did not knew i had a typo, but i just asked, did not know that i took a day off on that day π
No hype!