Use cases · Mobile apps
Mobile app mockups with your AI agent
Updated
Describe the app, and your agent lays out its screens on a Saymock board at real phone sizes: a sign-up flow, a home screen, the states in between. Every screen is editable HTML + Tailwind, your team comments on it, and you can click through the flow in a device frame before anyone writes app code.
How it goes
- Pick the device. Frames come in phone sizes (iPhone 16, iPhone 16 Pro Max, iPhone SE, Pixel 9, Galaxy S24), tablets and desktops. Name one in your prompt.
- The first screen sets the system. The agent builds the status bar, header and tab bar once, as components, then reuses them on every other screen. Lists use realistic data, not “Item 1, Item 2”.
- Then the flow. Screens go side by side in flow order, with the states that matter (empty, loading, error) next to their screen. Links between screens, bottom sheets and transitions make it a prototype you present in a device frame.
- The agent checks every screen with a screenshot and a layout check (overflow, misalignment, missing icons) before moving on.
- Review and iterate. Your team comments on elements; ask for alternatives and they stay attached to the screen, where everyone previews them and picks one.
Prompts that work
Use Saymock to design the onboarding of a habit-tracking app: four iPhone 16 screens, one accent color, real copy. Link them so I can click through.
Add the empty, loading and error states of the “Orders” screen next to it.
Propose three directions for the home screen: one minimal, one card-based, one with a large chart.
From mockup to a native app
When the screens are approved, the agent reads them back with get_specs: an
auto-layout tree (stacks, spacing, alignment, fill or hug sizes) in points or dp, the
colors and text styles, icons with their closest SF Symbol and Material Symbol, lists with
their data, components with their props. That is what it needs to write SwiftUI, Jetpack
Compose, Flutter or React Native in your project. See design to code.
Questions
Which AI agent should I use?
Any agent that speaks MCP: Claude, Claude Code, ChatGPT, Cursor, Codex, Copilot in VS Code, Gemini CLI, Le Chat. The setup for each is on the agents page. Saymock never calls a model itself.
Are the mockups images?
No. Each screen is HTML with Tailwind classes on a board. You select an element and change its text, color or spacing by hand, and the agent edits the same markup.
Can I use my own brand?
Yes. Give the agent your colors and fonts, or let it read them from your codebase; logos and images are uploaded to the workspace, fonts too.
Can I click through the flow?
Yes. Screens the agent links together play as a prototype in a device frame, with their transitions, bottom sheets and overlays.
Complete examples, with the prompt the agent was given: a booking app and an onboarding flow, in the gallery.
Related: wireframes, design review, Saymock for Claude Code, for Cursor.