An AI-native creative workspace for turning ideas into structured visual workflows.
一个把创意变成结构化视觉工作流的 AI 原生创作工作台。
canvas is built around a simple belief: creative work gets easier when assets, prompts, tasks, and outputs live in one shared operating surface instead of being scattered across tools.
canvas 的核心判断很简单：当素材、提示词、任务和结果都被收进同一个工作台，创作工作就会比在一堆分散工具之间切换轻很多。
Instead of treating generation as a single prompt box, canvas treats it as a collaborative system:
canvas 不把生成理解成一个孤立的输入框，而是把它当成一个可协作、可复用、可追踪的系统：
- infinite-canvas style visual planning
以无限画布的方式组织创意与生成链路 - shared workspaces with role-based collaboration
支持团队 workspace 和角色权限协作 - reusable subject, scene, and instruction libraries
沉淀可复用的主体库、场景库和指令库 - task execution across text, image, and video workflows
承接文本、图片、视频等多模态任务运行 - result tracking, retries, and workspace-level operations
跟踪结果、失败重试和 workspace 级别任务管理
Most AI creation tools optimize for one-shot generation.
I’m more interested in the layer above that: how teams actually organize assets, iterate prompts, run tasks, and keep context alive across a full workflow.
大多数 AI 创作工具优化的是“一次生成”。
我更关心的是上面那一层：团队到底怎么组织素材、迭代提示词、发起任务，以及在完整工作流里保持上下文连续。
That is the product direction behind canvas.
这就是 canvas 这款产品的方向。
- workspace-aware task and asset management
面向 workspace 的任务与资产管理 - canvas nodes for planning, generation, and review
面向规划、生成和回看的画布节点系统 - library layers for reusable creative context
让创作上下文可复用的资源库层 - multi-provider AI workflow infrastructure
面向多供应商的 AI 工作流基础设施 - session, permissions, and operational product scaffolding
真实 session、权限与产品化运行骨架
- Next.js
- TypeScript
- React
- Drizzle ORM
- PostgreSQL
- Tailwind CSS
npm install
npm run dev