Stuck on a product or your workspace? Tell us and we'll work through it with you.
Include your setup and any error messages so we can help faster.
Answers about the Space Layer, deeepCLI, Figma Plugin, Web Space, and deeepSpace MCP.
deeepSpace is the shared system layer that connects your design system and your UI codebase. In modern product teams, design and code evolve in parallel but rarely in lockstep, causing silent drift. The Space Layer bridges both realities by establishing explicit component-level mappings, tracking independent version histories, detecting structural, styling, token, variant, and visual differences, and giving designers, developers, and AI agents one shared source of truth.
Connection happens in three steps: First, designers use the Figma Plugin to selectively sync components along with their variants, properties, styles, and tokens. Second, developers run deeepCLI in their repository to analyze UI components, props, styles, and Storybook configurations. Third, in Web Space, teams map corresponding design and code components together, creating a persistent, component-level connection that powers version history, comparison, and drift detection.
deeepCLI is the developer-side connection to the Space Layer. It analyzes your existing UI codebase without refactoring, annotations, or a particular folder structure: running deeep init only adds a small deeep.json config file. It discovers reusable components, inspects props, styles, token references, and component hierarchy, and detects Storybook stories and configurations. It then normalizes this data into a consistent model and syncs it with deeepSpace so it can be mapped to design components.
You do not need to import design files wholesale. The deeepSpace Figma Plugin is intentionally selective and designer-controlled. You choose exactly which components or component sets belong in deeepSpace, and they sync with their structure, variants, and properties. Your file's styles and variables come along as tokens, while unrelated frames and scratch canvas content are never pulled in.
deeepSpace renders each mapped component for every design variant and compares it with the design. It reports structural differences (elements missing from code or only in code), styling and design token discrepancies (such as divergent radii, colors, or spacing, or a raw value where the design uses a token), and visual differences, shown as a pixel diff of the rendered component against the design. It also shows which design properties aren't mapped to a code prop yet, and which variants couldn't be compared.
Visual verification compares rendered code components against their corresponding design visuals rather than relying solely on static code analysis or manual inspection. deeepCLI builds a preview of your components with your project's own setup and renders each one with the props of every design variant. Storybook isn't required: when your project has it, deeepCLI uses your stories for realistic example data, and falls back to your Storybook build if the preview can't be built.
Yes, in beta. Compared components and their variants are also drawn at your project's tablet and phone sizes, exactly as your code is, and deeepSpace reports what breaks there: a component wider than the screen, a page that scrolls sideways, content spilling out of its box or overlapping, and softer notes such as tap targets under 24px. You don't need a mobile design for this, and it never changes the design-code score. When a design is made for a phone or tablet (a Breakpoint property or a name like "Top Nav / Mobile"), it is also compared with your code at that screen size.
deeepSpace follows a 'connected, not collapsed' model. Design and code remain distinct sources with independent release cycles and responsibilities. deeepSpace keeps a version history on both sides: design versions with what changed and who pushed them, and code versions with their commit, author, and message. When drift occurs, each difference is traced back to the design version or commit where it started, so teams can see when and where design and implementation diverged, and leave a note when a difference is intended.
deeepSpace MCP is the deeepSpace Model Context Protocol (MCP) server that connects your external AI coding assistants—such as Cursor, Claude Code, Windsurf, and custom LLM agents—to the Space Layer. Over HTTP, AI agents get access to your components, design specifications, tokens, design ↔ code mappings, and the latest comparison results (detected differences per variant), so they generate code that adheres to your actual design system.
deeepCLI supports React and Next.js, Vue 3 and Nuxt 3 and 4, and Svelte 5, in JavaScript or TypeScript, with or without Storybook. React projects built with Vite, webpack, or Create React App are supported; for visual comparison, deeepCLI builds them with your Vite config when you have one, and with its own Vite setup otherwise. Vue and Svelte projects need Vite (or Nuxt). In Nuxt projects, visual comparison builds your components with your project's own Nuxt, modules and plugins included. It normalizes styling across CSS, SCSS, LESS, and Tailwind CSS, and integrates with Storybook to capture configured component states and variants.
Web Space serves as the central hub where teams manage design systems across organized workspaces and projects. You can keep multiple design systems, products, or client brands in separate workspaces, each with its own members, component mappings, version histories, and comparisons. deeepSpace MCP signs in as you, so your AI agents can read the workspaces you belong to and nothing else. Each workspace has its own Crew subscription, billed per person in it, while projects inside a workspace are unlimited.
No. deeepSpace operates as a non-destructive observability, mapping, and alignment layer. The Figma Plugin reads the components you select and never changes your designs; it only stores a hidden link to your deeepSpace project in the file. deeepCLI never edits your source code: the only files it adds are its own deeep.json config and a component index. Any code suggestions generated by connected AI tools remain entirely under your team's review and control.