# Build products in the open Enhance gives your agent a real visual canvas—not a screenshot pasted into a chat. It can inspect product context, author editable screens, publish working prototypes, act on feedback, and verify the pixels with you. ## Start with the product, not setup 1. [Connect your agent](/docs/getting-started). 2. Open the populated example canvas Enhance creates for you. 3. Ask for a small change, a new screen, or a prototype review. 4. Watch the work appear on the canvas and inspect the result together. > The normal experience is browser sign-in and a user-scoped plugin. You do not need to install a CLI, copy a token, edit a project file, or understand MCP transports. ## Four complete workflows | Goal | What the agent completes | | --- | --- | | Design on canvas | Reads the real product language, creates editable native objects, and inspects a full render. | | Publish a prototype | Packages the build you selected, publishes it, exercises the result, and checks the pixels. | | Respond to feedback | Acts on clear comments, leaves ambiguity open, republishes where needed, and verifies before resolving. | | Apply a canvas design | Changes the owning product source, rebuilds, republishes, and compares it with the approved design. | ## Human and agent representations Every page is available as polished HTML and clean Markdown. Start with [llms.txt](/docs/llms.txt) for the compact agent index or [llms-full.txt](/docs/llms-full.txt) for one complete corpus.