Install Command — Wire Your Brand into AI Coding Tools
Copy a single install command from your Brand Stack and paste it into Claude Code, Cursor, or Codex to automatically wire your brand's colors, fonts, voice, and personas into any project.
Every Brand Stack includes a one-click Install Command — a ready-to-paste prompt that installs your complete brand bundle into Claude Code, Cursor, Codex, or any other AI coding agent. Once installed, your AI tools write, design, and code in your brand’s voice and visual system automatically.
What the install command does
When you paste the install command into an AI coding agent, it:
- Downloads your brand bundle via a secure, token-authenticated URL — no sign-in needed for the agent
- Wires color tokens into your project as CSS variables or design tokens in your framework’s format
- Sets up fonts using the
fonts.cssfile included in your bundle (Google Fonts@import+ CSS variable definitions) - Places logos and favicons in the correct asset directories
- Configures voice rules so the AI applies your brand’s writing style to all generated copy
- Loads persona context so the AI knows your target audience when writing for or about your customers
- Points to your live MCP via
mcp-pointer.jsonfor real-time brand queries during development
The agent reads an INSTALL.md guide that ships inside your bundle. It walks through each step in order, so the installation is consistent whether you’re using Claude Code, Cursor, or any other tool that reads markdown instructions.
How to copy your install command
- Open any Brand Stack in your workspace
- Find the install prompt at the top of the Build panel — the right-hand column of the detail page (it stacks below the sections on smaller screens)
- Click the copy button on the prompt preview
- Paste it into your AI coding tool’s chat
The command includes a secure URL pointing to your brand bundle. The agent fetches the bundle, reads the install guide, and applies your brand automatically.
Share tokens
Each install command includes a unique share token — a secure, revocable link to your brand bundle. The token grants read-only access to the bundle files; it cannot modify your stack or access any other workspace data.
To regenerate your share token:
- Open your stack’s detail page
- Click Rotate link in the Build panel’s header (the right-hand column)
- Your new install command is generated immediately
Any previous install command using the old token will stop working. Regenerate if you’ve shared your command somewhere you no longer control.
What’s in the bundle
The brand bundle your install command downloads includes:
| File | What it contains |
|---|---|
brand/visual/colors/tokens.css |
CSS custom properties for your full color palette |
brand/visual/fonts/fonts.css |
Google Fonts @import + CSS variable definitions for your type pairings |
brand/visual/logos/ |
Brand mark, wordmark, monogram in SVG and PNG, plus favicon set |
brand/voice/voice.md |
Voice descriptors, tone guide, vocabulary, writing samples |
brand/identity/identity.md |
Brand name, one-liner, mission, positioning, and about copy |
brand/personas/ |
One markdown file per persona with goals, language, and context |
mcp-pointer.json |
Pointer to the live Brand Stack MCP for real-time queries |
INSTALL.md |
Agent-readable step-by-step install guide |
Target profiles
Every install compiles the color and typography tokens to fit the codebase they’re landing in, instead of shipping one generic file. When an AI coding agent runs the install (directly, or via adopt_repo), Brand Stack detects the repo’s stack and picks a target profile automatically:
| Profile | What you get |
|---|---|
| Next.js + Tailwind v4 | globals.css with the @theme inline mapping — no config file needed |
| Next.js + Tailwind v3 | tailwind.config.ts extending theme.colors with your tokens |
| Tailwind v4 (framework-agnostic) | Same CSS-first output, generic asset placement |
| Tailwind v3 (framework-agnostic) | Same config-file output, generic asset placement |
| Plain CSS variables | A self-contained variables.css — no framework assumed |
Design tokens also ship as tokens.json in the industry-standard DTCG format alongside the CSS, so design tools that consume DTCG tokens can read your Brand Stack directly.
Pinning a version
Every install can target a specific, reproducible version of your brand instead of always pulling the latest. Pass a version like 1.2.0 when generating the install command (or as the pin input to the get_install_command MCP tool) to lock a repo to that exact snapshot — useful for a production deploy you don’t want shifting under you when the brand changes later. See Version History, Proposals & Connections for how versions are cut.
Using the MCP for real-time brand queries
The mcp-pointer.json file in your bundle points to Brand Stack’s MCP server. After running the install, your AI tools can query your live Brand Stack for specific brand data during development — persona details, color values, voice rules — without needing to re-run the install every time your brand is updated.
See MCP setup → for instructions on connecting your tools to the live MCP.
Which AI coding tools are supported
The install command is written for agent-style AI tools that read and execute markdown instructions:
- Claude Code — works with the
claudeCLI; paste the command into chat - Cursor — paste into the Composer or chat
- GitHub Copilot / Codex — paste as an agent instruction
- Windsurf — paste into the AI chat
- Any tool that can read file contents and run agentic tasks