Prompting

Describe what you want to build — in Hope AI, or in your own AI coding agent connected to Bit Cloud — and it is designed and implemented as composable components. The more context you provide, the better the result.

What you can build

Websites & landing pagesMarketing sites, portfolios, and multi-page websites.

Web applicationsInteractive apps with routing, state, and data fetching.

Design systemsThemed component libraries with tokens and documentation.

Backend servicesGraphQL and RESTful APIs with database support.

UI componentsReusable buttons, forms, cards, and layouts.

Full-stack platformsEnd-to-end systems combining UI, APIs, and data.

Attach

Writing effective prompts

Be specific about what you want. Name the pages, features, and interactions you need.
Describe the user experience. Explain how users should interact with your product.
Include design direction. Mention themes, colors, or attach a Figma file for reference.
Override defaults when needed. Hope AI uses best practices and proven tech stacks by default. Only specify technologies when you want something different.
Start broad, then refine. Begin with the overall vision. Refine individual components after generation.

Example prompts

Website

"Create a personal portfolio website with a hero section, project gallery with filtering, an about page, and a contact form. Use a dark theme with purple accents."

Design system

"Build a design system with Button, Input, Select, Card, and Modal components. Include light and dark themes, consistent spacing tokens, and accessibility support."

Full-stack app

"Build a task management app with user authentication, project boards, drag-and-drop task cards, and a backend API with database persistence."

Backend service

"Create an API for a blog platform with endpoints for posts, comments, and categories. Include pagination, search, and authentication."

Adding context

Figma designsConnect your Figma account and attach design files so Hope AI matches your visual design — layout, colors, typography, and spacing.

GitHub repositoriesConnect your Git provider and paste a repository URL to give Hope AI context from your existing codebase, patterns, and conventions.Connect Git provider

File attachmentsUpload images, screenshots, documents, or any other files to provide additional context for your prompt.

Plan mode

You can enter Plan mode explicitly by prompting "Plan a..." — a feature, platform, app, design system, and more. Hope AI also suggests Plan mode after generating a prototype — proposing the complete architecture and component breakdown for the full solution. Review and refine the plan before Hope AI builds it out.

What happens next

By default, Hope AI starts generating components right away. For complex solutions, it first generates a prototype to help you validate the direction before committing to the full build.

Using your own AI agent

Prefer your own agent? Prompt Claude Code, Cursor, Codex, or any MCP-capable agent the same way — from your editor or from a Workspace in the browser. Connected to Bit Cloud, it reuses the components your team already has and builds new ones as Bit components — ready for the same review, release, and deploy flow.

Connect your agentAdd the Bit Cloud MCP server to Claude Code, Cursor, Codex, Windsurf, or GitHub Copilot. Your agent can then search, inspect, and reuse your components.Set up MCP

Run your agent in a WorkspaceOpen a Workspace — a full VS Code in the browser with Bit and your agent’s connection to Bit Cloud already set up — and prompt your agent right next to your components. Nothing to install or configure.Start on Workspaces

Add agent instructionsShip instructions in your project so the agent looks for existing components first and builds new ones as Bit components.Agent instructions

Share conventions with AGENTS.mdPut your stack’s conventions in an AGENTS.md file. Hope AI and your own agents follow the same rules.AGENTS.md