Open Design

Powerformer, Inc.·Powerformer.OpenDesign

The open-source alternative to Claude Design. Your existing coding agent — Claude · Codex · Cursor · Gemini · OpenCode · Qwen — becomes the design engine, driven by 31 composable skills and 72 brand-grade design systems.

The open-source alternative to Claude Design. Local-first, web-deployable, BYOK at every layer — 16 coding-agent CLIs auto-detected on your PATH (Claude Code, Codex, Devin for Terminal, Cursor Agent, Gemini CLI, OpenCode, Qwen, Qoder CLI, GitHub Copilot CLI, Hermes, Kimi, Pi, Kiro, Kilo, Mistral Vibe, DeepSeek TUI) become the design engine, driven by 31 composable Skills and 72 brand-grade Design Systems. No CLI? An OpenAI-compatible BYOK proxy is the same loop minus the spawn. Why this exists Anthropic's Claude Design (released 2026-04-17, Opus 4.7) showed what happens when an LLM stops writing prose and starts shipping design artifacts. It went viral — and stayed closed-source, paid-only, cloud-only, locked to Anthropic's model and Anthropic's skills. There is no checkout, no self-host, no Vercel deploy, no swap-in-your-own-agent. Open Design (OD) is the open-source alternative. Same loop, same artifact-first mental model, none of the lock-in. We don't ship an agent — the strongest coding agents already live on your laptop. We wire them into a skill-driven design workflow that runs locally with pnpm tools-dev, can deploy the web layer to Vercel, and stays BYOK at every layer. Type make me a magazine-style pitch deck for our seed round. The interactive question form pops up before the model improvises a single pixel. The agent picks one of five curated visual directions. A live TodoWrite plan streams into the UI. The daemon builds a real on-disk project folder with a seed template, layout library, and self-check checklist. The agent reads them — pre-flight enforced — runs a five-dimensional critique against its own output, and emits a single <artifact> that renders in a sandboxed iframe seconds later. That's not "AI tries to design something". That's an AI that has been trained, by the prompt stack, to behave like a senior designer with a working filesystem, a deterministic palette library, and a checklist culture — exactly the bar Claude Design set, but open and yours. OD stands on four open-source shoulders: - alchaincyf/huashu-design — the design-philosophy compass. Junior-Designer workflow, the 5-step brand-asset protocol, the anti-AI-slop checklist, the 5-dimensional self-critique, and the "5 schools × 20 design philosophies" idea behind our direction picker — all distilled into apps/daemon/src/prompts/discovery.ts. - op7418/guizang-ppt-skill — the deck mode. Bundled verbatim under skills/guizang-ppt/ with original LICENSE preserved; magazine-style layouts, WebGL hero, P0/P1/P2 checklists. - OpenCoworkAI/open-codesign — the UX north star and our closest peer. The first open-source Claude-Design alternative. We borrow its streaming-artifact loop, its sandboxed-iframe preview pattern (vendored React 18 + Babel), its live agent panel (todos + tool calls + interruptible generation), and its five-format export list (HTML / PDF / PPTX / ZIP / Markdown). We deliberately diverge on form factor — they are a desktop Electron app bundling pi-ai; we are a web app + local daemon that delegates to your existing CLI. - multica-ai/multica — the daemon-and-runtime architecture. PATH-scan agent detection, the local daemon as the only privileged process, the agent-as-teammate worldview.

winget install --id Powerformer.OpenDesign --exact --source winget

Latest 0.23.0·September 20, 2026

Release Notes

🌟 Codename: Home, Rebuilt. 🧵 73 PRs · 8 contributors · 10 days — Open Design 0.23.0 rebuilds the place where work begins and continues. Home, recent projects, run status, and the project conversation now form one clearer route into the work instead of separate places to navigate. 🚀 🔥 Highlights

  • 🧠 A clarification can survive the real world. Ask Open Design to continue a task, restart while it needs an answer, then come back and answer it: OD Next keeps the thread resumable instead of quietly abandoning the work. Failed runs now tell the truth before finalization, while delivered prototype pages stay delivered—not mysteriously marked as failures. (#7575, #7980, #7942, #7931, #7949) Thanks @lorenzozanee, @Siri-Ray.
  • 💬 The conversation no longer drops its own evidence. Earlier question forms, repeated-file read ranges, artifact updates and successful image next steps stay attached to the right turn. Workspace writes stop posing as deliverables, and quiet auxiliary cards step out of the way. (#7996, #7997, #8019, #8047, #8057, #7920, #7956) Thanks @lefarcen.
  • 🔁 Retries finally pick up where you left them. Cloud retries wait for the authority and history they need; transient reads recover; folded tasks retain their blocked state and retry the actual failed run. A pending clarification also stays in charge until you answer it. (#7953, #7954, #7955, #7967, #7981, #7989) Thanks @app/open-design-crew.
  • 🛠️ ACP results remain readable—even when a run does not. Tool rows now appear while Codex or OpenCode is still preparing their arguments; buffered words, failed tool results and a real cancellation all make it back into the conversation, while URLs stop masquerading as filenames in the record. (#8117, #7959, #8015, #8046, #8063) Thanks @Siri-Ray.
  • 🖼️ Artifact-heavy chats stay in the order you chose. Cover-image failures have a useful fallback; live updates refresh the card you are already reading; image strips scroll; requested file ranges and selected artifact order remain visible. (#8045, #8060, #8062, #8013) Thanks @app/open-design-crew.
  • 🧰 Bring your own model without opening a privacy detour. Image input works when a BYOK provider cannot advertise every capability, while proxy Basic credentials stay out of export and media tool-token lanes. (#8035, #7999, #8006) Thanks @Siri-Ray, @readwrightexecute.
  • 📦 The installed app holds onto its engine. Packaged MCP and sidecar delivery remain intact, and updater cleanup no longer removes the launcher payload that is still doing the work. (#8027, #7963) Thanks @PerishCode.
  • 🏠 Home starts with a clear place to create. A fresh Home opens ready for a Prototype, and switching creation types keeps the thought already in the composer. An empty balance is answered before leaving Home, while the project hand-off holds its layout steady instead of sliding into place. (#8208, #8278, #8240, #8229) Thanks @Siri-Ray.
  • 🧩 Long-running work no longer replays itself. A paused clarification keeps its history and waits for you; a compaction continuation can finish committed work in the same session instead of repeating it. (#8225, #8250, #7958) Thanks @lefarcen, @Siri-Ray. 📥 Download: Tag open-design-v0.23.0.
Platform Architecture Assets
macOS Apple silicon DMG · SHA-256
macOS Intel DMG · SHA-256
Windows x64 Installer · SHA-256
🔁 Changed
  • Run clocks now begin when the run really begins; folded runs keep their own clocks, the composer stays visibly preparing during send admission, and retry cards describe the send gate they actually face. (#7921, #7927, #8059, #8069) Thanks @lefarcen.
  • Cloud and AMR account, balance and upgrade links now lead to the current dashboard routes instead of sending you on a dead-end detour. (#8126) Thanks @nettee.
  • Agent-owned Codex threads are archived after a terminal completion, keeping finished local work from lingering as if it were still alive. (#8051)
  • Client campaigns now keep an authorized placement visible through a temporary refresh problem, without waking the app solely to check it. (#7986, #8269) Thanks @alchemistklk. 🐛 Fixed 💬 Conversations and interface
  • Continue stays on the conversation turn that owns it; question controls follow the UI language; long question titles remain readable; and native tooltips no longer interrupt the choice flow. (#8014, #8010, #8031, #8033) Thanks @lefarcen.
  • Narrow chat panes keep footer actions and time visible; terminal artifact recovery respects the preview you chose; and uploaded attachments retain their true size. (#8001, #8050, #8030) Thanks @lefarcen.
  • A project-switcher menu stays usable while a run scrolls; queued messages show their drop point between cards; and the chat log no longer loses pointer hit-testing near its top edge. (#8223, #8228) Thanks @Siri-Ray, @lefarcen.
  • Campaign Test modals survive a same-deployment reload and can recover their replaced context instead of failing every placement. (#8210, #8263) Thanks @alchemistklk. 🔑 Models, media and diagnostics
  • AMR media-model chat now respects the model guard, and workspace diagnostics retain bounded environmental evidence when a failure needs explaining. (#7695, #7951) Thanks @Siri-Ray, @lefarcen.
  • Client safety telemetry carries platform context and uses canonical monitoring kinds, making real-world errors easier to find and compare. (#7944, #7945) 🙏 Thanks to everyone who shipped 0.23.0 @alchemistklk · @lefarcen · @lorenzozanee · @nettee · @PerishCode · @QiuQiuNB666 · @readwrightexecute · @Siri-Ray

Installer type: nullsoft

x64—FC0197E959F55CBA775112D71C722182988E80F007191B0EC08FB8C6BC79C0FE

Details

Homepage
https://open-design.ai/
License
Apache-2.0
Publisher
Powerformer, Inc.
Support
https://github.com/nexu-io/open-design/issues
Copyright
Copyright 2026 Open Design contributors

Tags

agentagenticaiclaudedesignlarge-language-modelllm

Older versions (29)

0.22.2
x64—7916EA5FBB4E61EDB186951AC3689F8B0ECCAC4F448F740C80D26067094CE0F2
0.22.1
x64—5E56D49C3839C35C2AB276FBEABBCB85D58952119D3E5B33D4FC877448759CFA
0.22.0
x64—9D9413F246D5BF36FFC581C121F6EA3EB44452681574DA84669A521437E6D1EE
0.21.1
x64—CD203B1C931FE1F7621929945B5AA8D2387A5C6FA5273CF66325A0B443224E7E
0.21.0
x64—E5ACAD0238EF5C2D77353DFEB5D4CBF3C4ED9E313ADEBCDFA54F55129C8F4269
0.20.2
x64—AEA88B7E089C7FA41E09CC56AA9C7D02380B2756EC207BD92BE90066B0640B94
0.20.1
x64—93E082779F43F7E807A2DD2AD553854921044F11796701FC42ECCA270AB7FDA3
0.20.0
x64—9788A6A7A3FF1C6E31B41756FB1ECD631B54307ACFED5937BFE5B1BD0000D9B8
0.19.2
x64—DD3A78A5ED73B6D7DA2E374E73A019470A00ABCE2F4F95896EB7B899C1DAD756
0.19.1
x64—E994FC6A707135092456E48366847AB95F0E96C72AD8A7F3574F6385616A6FE1
0.19.0
x64—D253518D29F44758FA200A2BF40589F3EE81A20ACC8ED5E3F2239FDAFA0BECF1
0.18.1
x64—DA28D803B31888861EE329667ECEB8454F27B33E0DDDE5C83782C150499C71D0
0.18.0
x64—A76A4C28016DE2239B864F0E1A864933BEA211C5ACA638FC3B27BD2AC02F48CB
0.17.0
x64—822924DF6728125A9744B99FA518F0C15842E7C7FFF92CEB83C14AF73BAECAC1
0.16.1
x64—8D6712A522517DE9FF7CA996645442F4E23E4DF6F1E95681C8603D94BA5BF7EA
0.16.0
x64—99DB28BBB4918954030AEB30412252E198ACDFCCD4880E1684D7ACF8C55EC698
0.15.1
x64—B83A32A7D6EC7FED96F383BB9A5DCD1C8DCC2C0FD0FA84E84882A9665D92449E
0.15.0
x64—63FC2E609489474E99187CDF94D01D063C1DBEE733AAF2464D835CDC1E96F6B5
0.14.1
x64—18D7F5243BBACC48DB9FF13EA1298FEB15142EE55C6296DF555A036B7B8C2EC2
0.14.0
x64—9A735868E65EDC16930615C5D3FD7CFF2A26EEF4533C3E45E9CCC4B35077FDFC
0.13.0
x64—68900AC273684A8362A007071C30FC06722ECDAA3C5490C6732D94E3476444AC
0.12.0
x64—736572D7B2AD59FE9CFC9FBF8E52D876D1F8F26990122F59A904957A4F770C7A
0.11.0
x64—4F27F75CD3EA012B7E4A0F5C72EBA102943331446282C4ECDE1129AECD532366
0.10.1
x64—E2837BFDBE115ECBFA1FAB711DED3B626142760A7E66CDB0F7FFA14F80E74381
0.10.0
x64—648DE15BEEB949B4664855199585F2434157D08E437927CB2BA3F53E6DFF9749
0.9.0
x64—AC08DF5B442C73E2847BFE94C843C3691AFA631CDA00D1407E008DC95E651235
0.8.0
x64—067E163F4BB439FF532CB656D3AD0DEAECF138B046650600454FC6920117D8B3
0.7.0
x64—5A77747DCDB17443EF880291B4166147AF53A96AAE1416BEEEF962E8ADE2DDD0
0.5.0
x64—66D81372FFB28936C1CB0001150DC295B7D817C2701BA86EABA2A282BD815131