Byterover - Memory layer for your AI coding agents
by•
Byterover is a self-improving memory layer for your AI coding agents—create, retrieve, manage vibe-coding best practices across projects and teams. You can start now by installing Byterover's extension via your AI IDE like Cursor, Windsurf, and more.
Replies
Byterover
MindPal
@andy_byterover congrats for the launch
Byterover
@maiquangtuan thank tuan; it would be grate if mindpal team can try and leave us your feedback
@andy_byterover Nice products bro, congrats for the launch!
Byterover
@tonyhothu Thanks so much, Tony — your support really means a lot to us!
Byterover
Hi builders everywhere, we can’t wait to see what you all do with our memory layer!
We’ve launched an earlier version. From Solana trading bots to automated Meta Ads tools, we’re seeing builders use Byterover for a variety of use cases—not just to store coding practices, but increasingly to capture vertical business logic of the application as well. Some use us to switch seamlessly between Cursor and Windsurf, and others without losing context.
Looking forward to seeing what you can build with our memory.
Byterover
@minh_phan6 it is exactly what byterover is buit for, this is the way we can fuse our knowlege with the agent's capability, kind of "context" engineering
MindPal
Congrats on the launch guys!!
Byterover
@sylviangth thanhs Tham!
Auralix
Congrats on the launch! I'm trying it for myself but curious about how does Byterover handle conflicting coding practices when sharing memory across teams?
Byterover
@chanitypham Hey, thanks a lot for the awesome question! So, ByteRover is what we call an “agentic memory” — basically, there’s an internal agent that handles all the memory stuff for you. Here’s how it works: when it notices a new coding experience, it creates a memory for it. If it sees something needs to be updated with new concepts, it updates it. And if it finds a concept that clashes with an old one, it’ll delete the outdated memory and replace it with a fresh one. Hope that makes things a bit clearer!
Byterover
@aadarshkt Thanks for your question! To be honest, we don’t know the exact internal mechanism of ChatGPT’s memory, and whether they use dedicated agents for memory management. However, we’re continuously working to improve our own memory management process. Some areas we’re focusing on:
- Enriching memory with more context, for example by integrating with your MCP server.
- Adding a reflection process, where the agent regularly re-evaluates and updates stored memories.
- Continuously improving memory quality and relevance.
We’ll keep updating our approach and are open to exploring strategies used in other systems like ChatGPT as we learn more.
I would love of it had like a api version or something we could integrate into our own app as a memory layer would love of if it could happen.
Byterover
@its_abhinav_jangid Hey! Thanks for your question. We’re planning to add more connection interfaces in the next release (coming soon!) so you’ll be able to connect ByteRover to your application via API calls. Right now, we’re focusing on IDE support, so MCP is the current connection interface we’re using.
@andy_byterover Thanks anyway you could notify me for early access or when you release. Killed the launch btw 👏
Byterover
@its_abhinav_jangid I’ll let you know as soon as our upcoming release is available.
For early access, we’ll notify you about our open-source version — you’ll be able to connect it to your app via APIs or MCPs, whichever works best for you. Plus, you’ll have the flexibility to customize it to suit your needs. We’re aiming to release it soon, likely next week!
Byterover
@aadarshkt Hey, thank you for your interest! We're working at full speed to improve the memory agent and open source it soon. In the meantime, feel free to join our community here: https://discord.com/invite/UMRrpNjh5W — we'd love to chat!
YouMind
Congratulations on the launch! Byterover’s approach to persistent memory for AI coding agents is truly innovative. Do you see potential for expanding this memory layer to enhance team collaboration?
Byterover
@hi_caicai Hey, thanks! We already support team collaboration — you can create an organization then workspace, invite your team members, and start collaborating on the same memory workspace. (This is actually how we’re using ByteRover ourselves right now!)
Byterover
I believe Byterover is solving a real pain of almost all developers right now, and the movement of AI coding is just in the early phase.
This is massive. 🔥 Congrats on the launch, guys!
Byterover
@polsabandal Thanks! It would be great if you could try it out and leave us some feedback about the product. We’re rolling out lots of improvements and updates very soon!
Congrats on the team! I wonder if there will be bi-directional syncing between the memory feature of each IDE supported?
Byterover
@kynamng Hey Nam, thanks so much for your time! That’s a great suggestion, and we’ll definitely work on it — it could bring a lot of value to users. Here’s what we’re thinking: right now, the memory generated by Cursor is written as rules. We’ll have an MCP server hooked up to our extension that reads those rules and syncs them into ByteRover memory. We’ll keep experimenting with this — excited to see where it goes!
MagicSlides App
Congratulations on the launch duy, but isn't this something cursor is adding in built?
Byterover
@indianappguy Hey, thanks! So, Cursor’s memory is just one of their features — but for us, memory is the product. There are three main things that make us different from Cursor:
1. Under the hood, we use both a vector database and a graph database to store memories for your vibe coding agent. This means the agent can retrieve memory semantically, with extra structure and context captured from the graph.
2. Memory is fully managed by an agent, so you don’t have to worry about how it’s stored. Everything — creating, updating, deleting memory — is handled automatically by the agent (we call it “Cipher” internally).
3. Our memory system is IDE-agnostic, so if you switch from Cursor to another IDE, it’s plug-and-play — and it just works.