SI Agent News
DevelopingOpenAIOpen modelsAgents

Yugoo Design is an open-source plugin that makes AI agents ask design questions

A new open-source plugin has your AI agent suggest what to decide next and show choices in panels, so you stay in charge of the design before any code is written.

The short version

Why it matters

Much of the AI conversation is about getting finished results faster. Yugoo bets that people also need a clear way to shape the decisions behind those results. If you build software with an agent, it is a concrete example of keeping a human in the loop at the planning stage, though the evidence so far is the makers' own description.

A consultant, not just a code generator

As dev.to reported in a post by one of its makers, Jeffrey, Yugoo Design is an open-source plugin that turns an AI agent into what the team calls a proactive design consultant. Instead of waiting for instructions and handing back a finished product, the agent helps you work through the design and suggests what to look at next.

The problem the makers describe is familiar to anyone who has tried "vibe coding," meaning describing what you want in plain language and letting an AI build it. Results arrive fast, but you may not have had a real say in the details. Even a small to-do app has features, workflows, permissions and business rules, and these affect one another. Changing how a task is handed from one person to another can change who may edit it, what happens next and what appears on screen.

The makers argue that a chat of questions and answers struggles once many decisions are tangled together. It is hard to hold all those relationships in your head from a text conversation alone. Yugoo tries to make them visible.

Bundles, panels and Kits

The basic building block is called a Bundle. It has three parts: a skill, which is a written set of instructions telling the agent how to do a piece of work; a data format that defines what the work produces; and an interactive panel where you can inspect and adjust that output. There is one Bundle for each aspect of a design.

The panel suits the information. A workflow can appear as a diagram, options can sit side by side for comparison, and permissions can be reviewed in a table. You can edit a detail straight in the panel or talk it over with the agent. Once you save, the agent uses your decisions to revise the design.

Bundles are grouped into a Kit, which also carries a domain-specific "design mindset." That mindset lets the agent judge how complete the design is, spot missing information and recommend the next thing to explore. Your feedback shifts those recommendations. A choice in one panel might raise a question in another, or show you need to do some research first. The makers describe the whole thing as a loop: the agent proposes, you judge, and the agent revises.

A knowledge graph to catch contradictions

Parts of a design can each look fine while the links between them hide gaps or clashes. A change to a workflow, for example, might affect permissions, business rules and pages. To catch this, Yugoo automatically builds a knowledge graph, a map of how pieces of information and decisions connect across the project.

A background review uses that map to find conflicts and propose improvements. You decide whether to accept a suggestion, reject it or come back to it later. You can also open the graph and explore the relationships yourself. The design and its history are kept in project files, so you and the agent can pick up later from decisions you already made.

Beyond web apps, and what to expect

The makers say Yugoo is not meant only for vibe coding. A tool called Yugoo Studio lets people build and manage their own Bundles and Kits, mixing skills, data structures, panels and domain methods. Panels can be customized with declarative UI, a way of describing what a screen should show rather than programming every step, or with A2UI in hosts that support it. The Kit's mindset and its set of Bundles can also be changed to suit a field.

One Kit ships with the plugin by default: the Web App Kit, meant for product ideation and feature design before coding begins. You can start from a fresh app idea or from an existing product you want to rethink. The team says it would like to explore Kits for presentations, animation, game design and mathematical modeling with the community.

Yugoo runs in agent apps such as ChatGPT and DeepSeek Harness. The project is licensed under Apache 2.0, a permissive open-source license, and the website has setup guidance while the GitHub repository includes a walkthrough. The makers call it an early release they are still testing and improving.

How to read the claim

This is the makers' own description, so treat it as a pitch rather than an independent review. The source offers no test results, user numbers or comparisons with other approaches. Whether the panels and the proactive suggestions actually lead to better outcomes is not shown.

The underlying idea is still reasonable. If an agent is going to make many choices for you, it helps to see those choices laid out and to be able to change them before they harden into code. The practical question is whether the extra step feels like useful structure or like extra work.

What to watch

Try it

Before installing anything, list the features, permissions and rules of an app idea on one page and note where one decision affects another. Then see whether Yugoo's walkthrough on GitHub handles those links the way you would.

Sources and links

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

Listen to the AI Wire headlines

Listen to the AI Wire

Listen to our daily AI shows every morning: on your commute, in the shower, wherever you are

Listen to the daily shows

First reported by dev.to · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

More AI news