The author introduces Vellium, an open-source cross-platform desktop application for interacting with LLMs, featuring new desktop widgets and a visual interface for AI agents that support MCP servers and file manipulation.
Hey everyone, I’m building Vellium, an open-source cross-platform app for working with LLMs in desktop workflows. The goal is to make it easier to use different models for writing, automation, coding help, file-based workflows, and agent experiments from one interface. The latest update adds two larger features. The first one is desktop widgets. You can create a small interactive AI widget and place it on your desktop above other windows. It can react visually, expose a small hover interface, and let you send quick messages without opening the full app window. This part is still experimental, so I’m currently looking for feedback on whether this kind of lightweight desktop interaction is actually useful in daily LLM workflows. The second feature is Agents. The app now has an optional Agents tab, disabled by default, which can be enabled in settings. The idea is to provide a more visual interface for CLI-like agent workflows. Agents can read documents, inspect folders, run terminal commands, help with code, edit files, and use connected tools. Vellium also supports MCP servers. If you already have MCP servers connected in the app, you can attach them to agents and use them inside the same workflow. Apart from these bigger additions, I also fixed a lot of bugs and added smaller improvements. For example, chat mode now supports custom fields, so users can add or remove fields depending on their workflow. The project is still evolving, and some parts are experimental, but I’d appreciate feedback from people interested in open-source LLM apps, agent interfaces, MCP workflows, or desktop AI tooling. GitHub: [https://github.com/tg-prplx/vellium](https://github.com/tg-prplx/vellium)
A developer shares an open-source desktop AI agent client called KainClaw, built with Electron, supporting MCP, multiple AI providers, tools, and background tasks. The project is still early and seeks community feedback.
Open-LLM-VTuber is an open-source voice-interactive AI companion with a Live2D avatar, supporting real-time conversations and visual perception, fully operable offline.
A developer with 45 years of experience is building a local-first harness for LLMs with multi-agent logic, soon to be open-sourced on GitHub, and asks the community what features would improve their local LLM experience.
PixelClaw is a free, open-source LLM agent that combines conversational AI with image generation, editing, and audio tools in a Raylib-based drag-and-drop UI.
Andrew Ng discusses the rise of desktop AI agents and coding CLI tools, introduces the open-source OpenCoworker project, and examines agent harness designs where LLMs drive autonomous task execution.