# Cube Pets Office: An AI Agent OS That Turns a One-Sentence Idea Into a Complete Product Solution

> Explore Cube Pets Office—a revolutionary AI Agent OS that collaborates via 7 specialized AI roles to turn a simple idea into a complete product solution (including requirement documents, system architecture, and task planning) in 5 minutes, with full visualization, auditability, and exportability.

- 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm)
- 发布时间: 2026-05-24T15:15:32.000Z
- 最近活动: 2026-05-24T15:18:05.151Z
- 热度: 133.0
- 关键词: AI Agent, 产品预演, 自动化文档生成, 多Agent协作, 可视化工作流, Docker沙箱, SPEC模块化, 产品方案推演
- 页面链接: https://www.zingnex.cn/en/forum/thread/cube-pets-office-ai
- Canonical: https://www.zingnex.cn/forum/thread/cube-pets-office-ai
- Markdown 来源: floors_fallback

---

## [Introduction] Cube Pets Office: An AI Agent OS That Turns Ideas Into Complete Product Solutions in 5 Minutes

Cube Pets Office is a revolutionary AI Agent OS developed by xiaojilele-glitch from the OpenCroc team, released on GitHub in May 2026. It collaborates via 7 specialized AI roles to turn a one-sentence idea into a complete product solution (including requirement documents, system architecture, and task planning) in 5 minutes, with full visualization, auditability, and exportability. Its core goal is to let AI "preview" the entire product for you, solving the problem of low efficiency in early decision verification in traditional development processes.

## [Background] Pain Points of Traditional Development and Core Concept of Product Preview

In traditional development processes, it takes days to write PRDs, weeks to align requirements, and months to validate directions from idea to verification. Current AI programming tools mostly focus on code generation, but Cube Pets Office specializes in "product preview"—before investing in development, let AI deduce the complete product building process: goal clarification (supplementing ambiguities and success criteria), path planning (multiple routes and risk assessment), modular decomposition, document generation, and effect preview. This is like an experienced product team quickly completing idea verification.

## [Methodology] Specialized AI Role Collaboration and Visualized Technical Architecture

### FSD Fleet (7 AI Roles)
| Role | Responsibility |
|:--|:--|
| Planner | Break down goals into executable paths |
| Clarifier | Fill information gaps and eliminate ambiguities |
| Researcher | Collect background and verify hypotheses |
| Generator | Produce outputs like requirement documents |
| Operator | Execute code in Docker sandbox |
| Reviewer | Check quality and mark issues |
| Auditor | Maintain evidence chain and compliance records |

### Visualized Workflow
A 3D office scene displays the AI collaboration status, and the right-side workbench shows progress in real-time streaming. The entire process is auditable and replayable, maintaining a complete evidence chain and lineage DAG.

### Technical Architecture
Layered design: Entry Layer (Browser/Feishu Bot) → Frontend Layer (3D Scene/Cockpit) → Cube Brain (Workflow/Task Runtime) → Projection Layer → Intelligence Layer (Three-level Memory/RAG) → Trust Layer (Hash Chain Audit) → Execution Layer (Docker Sandbox) → Interoperability Layer (A2A Protocol).

## [Evidence] Real-World Application Cases and Developer Experience

### Application Cases
| Input Idea | Output Products |
|:--|:--|
| AI Comic Platform | 6 SPEC modules, content pipeline, monetization plan, architecture design |
| Permission Management SaaS | 8 SPEC modules, RBAC design, multi-tenant solution, API contract |
| Sentiment Analysis Tool | 5 SPEC modules, data pipeline, model selection, alert mechanism |
| Independent Developer Accounting App | 4 SPEC modules, local-first architecture, synchronization mechanism, privacy compliance |
| Enterprise Knowledge Base |7 SPEC modules, RAG pipeline, permission system, indexing strategy |

### Export and Deployment
- Preview results can be exported as Markdown, ZIP, or online links
- Quick start: `git clone ... && cd ... && pnpm install && pnpm run dev:all`
- Pure browser mode: `pnpm run dev:frontend` or visit GitHub Pages demo
- System requirements: Node.js 22+, pnpm, optional Docker

Codebase scale: 486,000 lines of TypeScript, 7771 test cases, 273 SPEC directories.

## [Conclusion & Outlook] Product Value, Limitations, and Future Directions

### Value
Cube Pets Office provides a low-cost trial-and-error tool for product managers, independent developers, and startup teams. It lets AI handle information collection and solution deduction, while humans focus on key decisions.

### Limitations
Currently in the Alpha stage, it focuses on product document and architecture deduction, with limited actual code generation capabilities. The collaboration efficiency of the 7 roles, cost overhead, and decision quality in complex scenarios need to be verified.

### Future Directions
It represents an important direction for AI as an intelligent collaborator in product teams. The goal is to optimize collaboration efficiency and solve the problem of "making better decisions before writing code".