Cipher is an open-source memory layer, connected to AI IDEs and CLIs through MCP. Auto-generate coding memories that scale with your code-base, auto-retrieve relevant coding memories and knowledge across IDEs, projects and teams.
Hi Product Hunt Community,
Having received positive feedback from the community since launching ByteRover, we have decided to open-source our core AI engine - memory layer.
This version Cipher got a major upgrade since latest version of Byterover, rolled out last month, with improved memory search tool, and faster memory creation. Especially this time, we open-source it so that developers can self-host, contribute and use it dynamically.
Key feature of Cipher:
🧠 Auto-generates coding memories that grow with your codebase
🔌 Works with any IDE — including modern models like QwenCode, Trae, Kimi K2, Kiro — not just VS Code, Cursor, or GitHub Copilot
🔁 Seamless memory + context switching across IDEs
👥 Share AI coding memories across your team in real time
🧬 Dual Memory Layer:
System 1: Programming concepts, business logic, past interactions
System 2: Reasoning steps behind the model’s code generation
⚡️ Zero-config installation on your IDE
With 1 month of work putting into this project, we hope this enables more developers and researchers to build advanced coding agents and drive innovation forward together.
Would love your feedback, contribution and support if ByteRover brings you value 🙌
Byterover acts like a memory layer for your best practices, think instant recall for vibe-coded patterns across projects and teammates. With plug-and-play extensions for IDEs like Cursor and Windsurf, it's the kind of upgrade that feels inevitable once you try it.
Love the open-core model you're using here, it's a great contribution. Just curious about the future: what's your general philosophy on how new features will be allocated between Cipher and Byterover? Thanks!
Wait, a memory layer for AI coding agents? That could be a total game changer—I’m always annoyed when bots “forget” earlier context. How does it handle long-term project details?
Amazing!!👏 Congratulations to the launch!🎉
This tool must has immense potential for streamlined development and enhanced AI capabilities. Kudos to the team for pushing the boundaries of innovation! 👏💻
Huge respect to the team — love the idea of memory layers that travel with the dev, across tools and teams. The System 1 / System 2 framing is particularly clever.
Quick question: as you scale adoption across larger teams or contributors, how are you thinking about support or real-time guidance? Especially when something breaks deep in context memory?
We’ve been thinking a lot about this space — launching soon on PH with Exthalpy, a live AI video agent to help users in those critical stuck moments. Would love your support when we go live 🙌
All the best,
Udit (fellow PH builder)
Impressive work! 🔥
Byterover adds real memory to AI coding agents, making them smarter across sessions and teams. The IDE integrations and open-source Cipher engone are impressive.
Excited to see how this evolves—especially for larger codebases and team workflows. Congrats on the launch!
Congratulations! I happened to be troubled by the problem that cloude code couldn't obtain the context of the cursor. It was very timely!
Huge congrats on launching ByteRover! Running something like Kimi K2 locally isn’t easy it usually means heavy infra and a steep DevOps curve. Would love to hear how you're approaching setup, scaling, and making it easier for folks who might not have access to high-end hardware
Congrats! Self-hosting an AI memory engine like Kimi K2 requires powerful hardware and DevOps skills. How do you plan to help users manage these high infrastructure demands and setup complexity
Nice work Duy, Cipher looks solid. We’re curating top dev tools at bestofweb .site, would be great to include ByteRover there too.
Interesting, isn't that is what Byterover already doing? Do I need to install Cipher if I already have Byterover?
Hi Product Hunt,
We are excited to contribute to community an opensource memory layer for coding agents.
It would be nice if we receive contribution from community to improve memory layer, because it helps vibe coders more efficiently.
Would love your feedback 🫶🏻
