nexu-io/open-design

nexu-io/open-design — 🎨 The open-source Claude Design alternative. 🖥️ Local-first desktop app. 🖼️ Your coding agent becomes the design engine:

Updated 2026-07-27 · Source https://github.com/nexu-io/open-design · 中文版

What is the idea?

A local-first desktop app that turns any of 20+ coding-agent CLIs (Claude Code, Codex, Cursor…) into a design engine: prototypes, dashboards, slides, images and video as real exportable files (HTML/PDF/PPTX/MP4), BYOK.

Who would use it, and what do they get?

Already consumer-facing; the unserved slice is the non-technical user who will never configure a CLI or an API key.

Why would anyone bother?

The person who needs a pitch deck, a floor plan or a listing sheet by tonight does not want a design tool — they want the file. This is the one shape of developer-facing product the board still allows: the AI hands back a finished artefact a human uses, not a part.

What is missing today?

Design tooling for people who can't design is still won by whoever packages agents best — but this project itself is filling that gap at speed, with 390 contributors and its own paid Cloud tier.

What makes this one different?

Not head-on. The open flank is vertical: one audience (e.g. teachers, real-estate agents, wedding planners) with templates, hosting and zero BYOK setup — the parts a horizontal open-source tool deliberately leaves out.

What is the sharpest thing about it?

Until this year an agent could write code but could not hand you a PPTX, a PDF or an MP4 you could open — you got a chat transcript and did the export yourself. Real file output, plugged into 20+ agent CLIs through one socket, is the thing that changed.

What could kill it?

You cannot out-build a 390-contributor project with a funded core team and its own monetization; any 'lite' clone starts three months behind an org shipping 515 commits a month.

Who pays, and for what?

Sell a hosted/team tier and enterprise licenses to design-dev teams who want agent-driven design without shipping their files to a cloud vendor.

What is the main risk?

Figma and the model vendors can bolt on the same agent-design flow; 'local-first' alone won't hold unless the agent-to-design UX is clearly better.

How much work is it?

not a build — logged as a capability signal

Where does the evidence come from?

https://github.com/nexu-io/open-design

GitHubWatching

Narcooo/inkos  ·  nashsu/llm_wiki