Arkloop – Open-source, local-first Agent client
Arkloop – Open-source, local-first Agent client
Hi HN, I built Arkloop – an open-source, local-first Agent client. You can think of it as Claude Desktop, but open source with its own taste. Not a wrapper around any Agent SDK, not based on Claude Code – I built this from scratch, solo, over 3 months. One thing I focused on: reducing cognitive load. For example – if you only use one model for chatting, why does every message show which model was used? That's noise. Another example: developers love showing full tool use details in the UI, but what users actually need to know is that the agent is working and which direction it's going. We don't need all that detail. Design philosophy: cognitive load, information density, value-driven, aesthetics. I'd love your feedback – on anything from a 1px spacing issue to a major architectural direction. The README has a section about this that I'm particularly proud of: "I welcome all forms of contribution. Even if you're not a developer – if something feels off during use, even just a spacing issue or a color, open an issue. If you're a developer, Arkloop's agent core, memory system, and hook mechanisms are all open." It also supports importing configs from OpenClaw/Hermes. GitHub: https://github.com/qqqqqf-q/arkloop Website & Download: https://arkloop.cn
Share cardActual performance
Launch Intel predictions
Analyze your own launch →Correct prediction on native model
Similar products
I made an open source and local translation app
Every 4s, Familiar OCRs my screen into Markdown (open source, local)
Open-source local voice typing
An open-source, local-first Webflow for your own app
An Open-Source, Local-First Agent Framework in Rust
ApiArk – Open-source, local-first API platform (Postman alternative)
Open-source, local-first GTD app — no account, no cloud
Open-source, local-first dictation you can trust
The open-source, local-first Figma for React apps
Open source, local voice-to-text