GitHub Copilot app for Beginners: How to build custom workflows with canvases
GitHub introduces customizable canvases in its Copilot app, allowing users to describe workflows in plain English to generate live, shared interfaces without coding, enabling real-time collaboration with AI agents.
GitHub Copilot’s new canvas feature lets users describe a desired workflow in plain English, and the AI agent generates a live, shared interface tailored to that workflow. This interface can take the form of a kanban board, checklist, dashboard, or other tool, eliminating the need to adapt existing fixed layouts. The canvas is bidirectional, allowing both the user and the agent to make updates simultaneously, similar to a shared whiteboard.
Creating a canvas requires no coding or design work. Users open an agent session, enter the /create-canvas skill, and describe their workflow. The agent then builds the interface in the right-side panel, ready for immediate use. The initial canvas serves as a starting point that can be refined over time by adding columns, filters, or integrating data like pull requests.
Once created, canvases are saved as extensions that can be reused or shared with teams. The shared state of the canvas updates in real time, allowing both the user and the agent to see changes instantly without manual syncing. This enables seamless collaboration, where commands or actions from either side appear immediately in the interface.
For those unsure where to start, the community has shared pre-built canvas extensions via Awesome Copilot, covering tools like release notes generators, kanban boards, and issue triage workflows. Users can install and customize these extensions to fit their specific needs, or create their own from scratch by describing a simple workflow.