How to Build a Brand Context Folder for Claude Code: Voice, Visual Identity, and Positioning
A brand context folder gives Claude your voice profile, design tokens, and ICP so every output inherits your identity. Here's how to build one in 30 minutes.
Why Claude Produces Generic Output (and How a Brand Context Folder Fixes It)
If you’ve used Claude for content creation and felt like the output was technically correct but somehow… not yours — you’re not imagining it. Claude is working with whatever context you give it. Without a clear picture of your brand, it fills the gaps with defaults: neutral tone, generic structure, safe positioning.
The fix isn’t a better prompt. It’s a better information architecture.
A brand context folder gives Claude a persistent reference point for everything that makes your brand yours — your voice profile, your design tokens, your ideal customer profile (ICP), and your positioning. Once it’s in place, every piece of Claude generates inherits your identity instead of borrowing someone else’s.
This guide shows you how to build one in about 30 minutes.
What a Brand Context Folder Actually Is
A brand context folder is a structured directory of plain-text and markdown files that you pass to Claude at the start of a session — or reference in a system prompt. It’s not a pitch deck or a brand guide PDF. Those are built for humans. This is built for Claude.
The folder typically lives in your project root, sits alongside your codebase or content files, and gets loaded into context when you’re running Claude Code or a Claude-powered workflow.
Other agents start typing. Remy starts asking.
Scoping, trade-offs, edge cases — the real work. Before a line of code.
Think of it as a briefing packet. Before a freelancer starts work, you hand them documents covering your tone, your audience, your visual rules, and what makes you different from competitors. This folder does the same thing, but for an AI that can process it instantly and apply it consistently.
Why Plain Text and Markdown?
Claude reads plain text best. PDFs, slide decks, and design files require conversion and introduce formatting noise. Markdown files load cleanly, reference quickly, and work across editors, terminals, and AI interfaces without friction.
The folder structure also matters. Clear filenames like voice-profile.md, design-tokens.md, and icp.md tell Claude what each file contains before it reads a word.
Step 1: Build Your Voice Profile
The voice profile is the most important file in your folder. It defines how your brand sounds in writing — not just “professional” or “casual,” but the specific characteristics that distinguish your copy from everyone else’s.
Define Your Tone Dimensions
Start with three to five dimensions that describe your voice. For each one, define the dial — where you sit between two poles — and give examples.
## Voice Profile
**Tone dimensions:**
- Directness: Direct (8/10) — We say what we mean. No hedging.
- Warmth: Warm but not casual (6/10) — We respect the reader's intelligence.
- Formality: Low formality (3/10) — Contractions are fine. So are short sentences.
- Authority: High (8/10) — We speak from expertise, not opinion.Numbers on a scale aren’t scientifically precise, but they give Claude a relative sense of positioning that’s more useful than abstract adjectives alone.
Include “We Say / We Don’t Say” Examples
This is where most brand guides fall short. Generic guidance like “be conversational” is hard to apply consistently. Specific examples are much easier.
## What We Say vs. What We Don't Say
| We say | We don't say |
|--------|--------------|
| "Here's how it works" | "Let's explore this exciting opportunity" |
| "This takes about 15 minutes" | "Quick and easy setup" |
| "Most teams see results in a week" | "Dramatically accelerate your ROI" |
| "You can skip this if..." | "Note: please be advised that..." |Add Sentence-Level Style Rules
List any specific patterns Claude should follow or avoid:
## Style Rules
- Reading level: 8th–10th grade
- Sentence length: average 14–18 words. Mix short punchy sentences with longer ones.
- Paragraph length: 2–4 sentences maximum in body copy. One sentence is fine.
- Use active voice
- Avoid starting sentences with "However," or "Additionally,"
- Use second person (you/your) throughout
- Don't use exclamation marks except in UI microcopy
- Avoid passive constructions ("it was determined that...")Claude responds well to numerical constraints. “Short paragraphs” is vague. “2–4 sentences” is executable.
Provide Three to Five Sample Paragraphs
The clearest signal you can give Claude is an example of your actual writing. Pull excerpts from content you’re proud of — a blog post, an email, a product page — and paste them in directly.
Label them clearly:
## Voice Examples (use these as style reference)
**Example 1 — Product explainer:**
[paste excerpt]
**Example 2 — Email to customers:**
[paste excerpt]Claude will pattern-match against these when generating new content, which is more effective than any description you could write.
Seven tools to build an app. Or just Remy.
Editor, preview, AI agents, deploy — all in one tab. Nothing to install.
Step 2: Document Your Visual Identity as Design Tokens
Most AI writing workflows skip visual identity entirely. That’s a mistake if you’re using Claude to generate anything that touches design — landing pages, component copy, social graphics, email templates.
Design tokens are the specific values that define your visual language: colors, typography, spacing, and UI patterns. When Claude knows these, it can generate HTML, CSS, or component code that matches your actual brand instead of making visual decisions on its own.
Color Tokens
## Design Tokens — Colors
**Primary palette:**
- Brand primary: #1A1A2E
- Brand accent: #E94560
- Background default: #F8F9FA
- Surface elevated: #FFFFFF
- Text primary: #1C1C1E
- Text secondary: #6B6B6B
**Semantic colors:**
- Success: #22C55E
- Warning: #F59E0B
- Error: #EF4444
- Info: #3B82F6Hex values alone still leave Claude guessing about where each color goes. Add usage rules underneath:
**Color usage rules:**
- Brand accent is for CTAs only. Never use it decoratively.
- Never place text directly on brand accent.
- Backgrounds stay light in onboarding and empty states.
- If you have a dark mode palette, document it separately with a --dark prefix.Typography Tokens
## Design Tokens — Typography
**Font families:**
- Headings: "Inter", sans-serif
- Body: "Inter", sans-serif
- Monospace: "JetBrains Mono", monospace
**Scale:**
- H1: 48px / 1.1 line-height / 700 weight
- H2: 36px / 1.2 line-height / 600 weight
- H3: 24px / 1.3 line-height / 600 weight
- Body: 16px / 1.6 line-height / 400 weight
- Caption: 12px / 1.5 line-height / 400 weightSpacing, Radius, and Component Patterns
If you’re using a component library or design system, include the core spacing scale and any component-level conventions:
## Design Tokens — Spacing, Radius, Shadows
Base unit: 4px
Scale: 4, 8, 12, 16, 24, 32, 48, 64, 96, 128
**Border radius:**
- Small: 4px (inputs, tags)
- Medium: 6px (buttons)
- Large: 8px (cards, modals)
- Full: 9999px (pills, avatars)
**Shadows:**
- sm: 0 1px 2px rgba(0,0,0,0.05)
- md: 0 4px 6px rgba(0,0,0,0.07)
- lg: 0 10px 15px rgba(0,0,0,0.10)
**Component conventions:**
- Cards: 24px padding, 8px border-radius, 1px border (#E5E7EB)
- Buttons: 12px vertical / 24px horizontal padding, 6px border-radius
- Section spacing: 96px vertical between major sectionsRadius and shadow values are the ones teams skip most often, and they’re usually why an AI-generated component feels subtly off sitting next to your real product.
When Claude Code generates a new component or layout, it can reference these tokens directly instead of guessing. The output lands much closer to production-ready on the first pass.
Logo and Asset Rules
Claude can’t see your logo, but it can follow rules about how to place and reference it in code and copy:
## Logo Usage
- Primary logo: horizontal lockup (mark + wordmark)
- Icon-only version: use when available width is under 120px
- Minimum clear space: equal to the height of the logo mark on all sides
- Approved backgrounds: white, brand primary, background default
- Never stretch, rotate, recolor, or place on a busy backgroundEveryone else built a construction worker.
We built the contractor.
One file at a time.
UI, API, database, deploy.
Step 3: Define Your ICP and Positioning
Positioning context helps Claude write for the right person in the right situation. Without it, Claude targets a generic professional — educated, busy, interested in your category, but not specifically your buyer.
Your Ideal Customer Profile
Write this in plain, specific language. Avoid aspirational descriptions (“our customers are visionaries who…”). Describe real people:
## Ideal Customer Profile
**Primary ICP — Operations Lead at a mid-market SaaS company**
Demographics:
- Title: Director of Operations, Head of RevOps, or similar
- Company size: 50–500 employees
- Industry: B2B SaaS, professional services, fintech
Situation:
- Managing 3–10 person team
- Owns tooling decisions but reports ROI upward
- Spends significant time on manual reporting and cross-tool coordination
Pain points:
- Data lives in five different tools and never syncs correctly
- Spends Friday afternoons cleaning spreadsheets instead of analyzing them
- Has tried point solutions that added complexity instead of reducing it
What they care about:
- Time to value (not features)
- Reliability and trust
- Whether they can explain this to their CFO in two sentencesThe more concrete the ICP, the better Claude can calibrate word choice, examples, objection handling, and the level of technical detail in any given piece.
Positioning Statement
Include a crisp positioning statement that Claude can use as a guiding constraint:
## Positioning
**For:** Operations and RevOps leads at mid-market B2B SaaS companies
**Who:** Spend too much time on data coordination instead of decision-making
**We are:** A workflow automation layer that connects your existing tools
**Unlike:** Point solutions that require rebuilding your stack
**We:** Sync, clean, and route your data automatically so your team can focus on the work that matters
**Key differentiators:**
1. Connects to your existing tools — no migration required
2. No-code setup — doesn't need engineering time
3. Visible logic — you can see exactly what's happening and whyIf you can’t fill this in cleanly, your positioning isn’t ready to hand to Claude yet — which is useful to find out now rather than three drafts in.
Messaging Hierarchy
Not every claim deserves equal prominence. Without a hierarchy, Claude will lead with whatever sounds strongest in isolation — often a supporting detail. Tell it what to open with and what to hold back:
## Messaging Hierarchy
**Tier 1 — lead with these:**
- Connects to your existing stack, no migration
- No-code setup, no engineering time required
- Visible logic — you can see what ran and why
**Tier 2 — supporting claims:**
- Most teams are live in under a day
- [N] prebuilt connectors
- Used by ops teams at [customer names]
**Tier 3 — mention only when relevant:**
- Free tier available
- Custom scripting for technical users
- SOC 2 Type II
**Claims to avoid:**
- Never say "the best" or "the only"
- Never name a competitor unless the reader does first
- Don't open with pricingThis is the file that stops Claude from writing a headline about your fourth-most-important feature.
Competitor Framing
Remy doesn't write the code. It manages the agents who do.
Remy runs the project. The specialists do the work. You work with the PM, not the implementers.
If there are common alternatives your audience considers, give Claude brief framing for how to position against them — without disparaging:
## Competitive Context
**vs. Zapier:** We're built for multi-step workflows that require logic and conditions, not simple linear triggers.
**vs. in-house scripts:** We maintain the infrastructure, handle errors, and give non-technical teammates visibility.
**vs. enterprise automation platforms:** We're faster to set up and don't require a dedicated implementation partner.Claude will use this when writing comparison content, objection-handling copy, or sales emails where competitive awareness matters.
Step 4: Add Supporting Reference Files
Beyond the three core files, a few additional documents significantly improve Claude’s output quality.
Glossary and Terminology
Every brand has its own vocabulary. Define it:
## Terminology
**Use these terms:**
- "Workflow" (not "automation" or "flow")
- "Connect" (not "integrate" or "sync")
- "Team" (not "users" or "customers" when writing to customers)
- "Source of truth" (acceptable phrase for our category)
**Avoid:**
- "AI-powered" (overused — describe the specific capability instead)
- "Best-in-class" (meaningless)
- "Seamless" (overused in our category)
- "Robust" (vague)Content Formats and Templates
If you generate recurring content types — weekly newsletters, LinkedIn posts, case study outlines — include the structural template for each:
## Content Templates
### Blog Post Structure
1. Lead with the problem (1–2 paragraphs)
2. State what the post covers and why it matters
3. Main body (H2 sections, 3–6 sections)
4. Practical takeaway or action step
5. Soft CTA
### LinkedIn Post Structure
- Hook: one-line insight or observation (no question)
- Context: 2–3 short paragraphs
- Takeaway or list
- No hashtags
- No emoji except for emphasis (max 1)Step 5: Structure the Folder
File organization matters because Claude processes context in order. Put your most important files first, and keep filenames self-explanatory.
Here’s a recommended structure:
brand-context/
├── README.md # Short overview and how to use this folder
├── voice-profile.md # Tone, style rules, examples
├── positioning.md # ICP, positioning statement, competitive framing
├── design-tokens.md # Colors, typography, spacing
├── terminology.md # Glossary, use/avoid terms
└── templates/
├── blog-post.md
├── linkedin-post.md
└── email-campaign.mdWrite a README for the Folder Itself
This might feel unnecessary, but it helps when you share the folder with teammates or when Claude needs to understand the purpose of each file quickly:
# Brand Context Folder
This folder contains reference documents for AI-assisted content creation.
Load these files into context at the start of any session where you're
generating brand content.
**File priority (load in this order):**
1. voice-profile.md — Most important. Always load.
2. positioning.md — Load for marketing copy, emails, and sales materials.
3. design-tokens.md — Load for any code or layout generation.
4. terminology.md — Load for any written content.
5. templates/ — Load the specific template for the content type you're creating.Step 6: Load It Into Claude Code
Once the folder is built, using it is straightforward.
Option 1: Reference It in Your System Prompt
If you’re building a Claude-powered workflow or agent, include instructions to read the brand context folder in the system prompt:
You are a content assistant for [Brand Name]. Before generating any content,
read and apply the documents in the /brand-context/ folder. Always follow
the voice profile, use terminology from the glossary, and write for the ICP
defined in positioning.md.Option 2: Use CLAUDE.md in Claude Code
Claude Code supports a CLAUDE.md file in the project root that automatically loads as context for every session. You can reference your brand context folder directly in this file:
## Brand Context
For all content generation, apply the brand guidelines in /brand-context/.
Key files:
- Voice and tone: /brand-context/voice-profile.md
- Positioning and ICP: /brand-context/positioning.md
- Design tokens: /brand-context/design-tokens.mdThis means every Claude Code session in that project starts with brand context already loaded — no manual prompting required.
Option 3: Load Files Explicitly in Each Session
For ad-hoc use, you can reference specific files at the start of a session:
Before we start, read these files:
- /brand-context/voice-profile.md
- /brand-context/positioning.md
Now help me write a product announcement email for [feature].Step 7: Calibrate It Before You Trust It
Building the folder gets you most of the way. The rest is one test pass and a maintenance habit.
Run a Calibration Prompt
Before you use the folder on real work, give Claude a small task it hasn’t seen — a product description, a two-paragraph email, a piece of UI microcopy — and put the output next to your writing samples.
Four questions:
- Does the tone match, or is it close but slightly off?
- Are the visual decisions consistent with the design tokens?
- Does it lead with a Tier 1 message?
- What feels wrong, and which file should have prevented it?
That last question is the whole exercise. Every miss points at a gap in a specific file.
Add Edge Case Rules as You Hit Them
Each time Claude produces something that feels off-brand, write the rule that would have caught it and add it to the relevant file. Over a few months the folder turns into a record of every brand decision your team has actually had to make.
The gaps that show up most often:
- Content types nobody documented — FAQs, error messages, notification copy
- Tone under pressure — how to write about a price increase or an outage
- Technical contexts — variable naming, code comment style, commit messages
Give It an Owner
Put the folder in version control and give one person responsibility for it. Review each file quarterly. If you’ve been maintaining it as you go, that review takes about 20 minutes.
Where MindStudio Fits in This Workflow
A brand context folder is powerful on its own, but it works even better when it’s part of a repeatable, automated content workflow rather than something you load manually each session.
MindStudio is a no-code platform for building AI agents that can wrap this entire process — loading your brand context, passing it to Claude, and generating structured outputs — into a workflow any teammate can run without touching the underlying files.
For example, you could build a MindStudio agent that:
- Accepts a content brief as input (content type, topic, target keyword, goal)
- Automatically loads the relevant brand context files from a connected source (Notion, Google Drive, or a file store)
- Passes that context to Claude alongside the brief
- Returns a brand-consistent draft formatted according to your template
Other agents ship a demo. Remy ships an app.
Real backend. Real database. Real auth. Real plumbing. Remy has it all.
The result is a repeatable content system that applies your brand rules every time — not just when a team member remembers to load the right files.
MindStudio supports Claude alongside 200+ other AI models, and connects to 1,000+ tools, so you can pipe outputs directly to your CMS, Slack, or wherever your team works. You can try it free at mindstudio.ai.
If you’re also building other AI-assisted workflows — image generation that follows your visual identity, or social content that matches your voice — MindStudio’s AI Media Workbench lets you chain media and text generation into a single automated pipeline.
Common Mistakes to Avoid
Keeping the Files Too Abstract
“Write with warmth” is less useful than “use contractions, write in second person, and avoid formal salutations.” Concrete guidance outperforms abstract direction every time.
Making the Folder Too Long
More context isn’t always better. Claude has a context window, and a bloated brand context folder competes with the actual task you’re asking it to complete. Keep each file focused. Aim for 300–600 words per file, not 3,000.
Skipping the Examples
Voice descriptions without examples are unreliable. Always include at least three real writing samples that demonstrate what your brand sounds like in practice. Claude will use these as anchors.
Covering Marketing Copy and Nothing Else
Most brand context folders document the blog and the landing page, then stop. But Claude also writes your error messages, UI strings, code comments, internal docs, and support replies — each with its own conventions. Add a short section for whichever of those your team actually produces.
Never Updating the Folder
Brand voice evolves. If your messaging shifts — a new ICP, repositioning, an acquired product line — update the folder. Stale context produces stale output.
Frequently Asked Questions
What is a brand context folder for Claude?
A brand context folder is a structured set of markdown files that defines your brand’s voice, visual identity, ideal customer profile, and positioning. You load these files into Claude’s context at the start of a session so it can generate content that matches your brand’s specific style — rather than defaulting to a generic tone.
How do I give Claude my brand voice?
The most effective approach combines three elements: a tone description with specific dial settings (e.g., directness 8/10), a list of “we say / we don’t say” word pairs, and actual writing samples from your best existing content. Examples carry more weight than descriptions alone.
What should I include in a Claude Code brand context?
At minimum: a voice profile (tone, style rules, writing samples), a positioning document (ICP, positioning statement, competitive framing), and a terminology glossary (preferred and avoided words). Add design tokens if you’re generating code or layouts, and content templates if you have recurring formats.
How do design tokens help Claude generate better content?
Design tokens give Claude the exact values — colors, font sizes, spacing — that define your visual identity. When Claude generates HTML, CSS, or component code without them, it makes visual decisions on its own that rarely match your actual brand. With tokens, the output aligns with your design system from the start.
Can I use a brand context folder with Claude Code?
Yes. Claude Code supports a CLAUDE.md file in the project root that loads automatically as persistent context for every session. You can reference your brand context folder in this file so Claude always starts with your brand guidelines in scope — without requiring manual loading each time. Note that only CLAUDE.md loads on its own; the other files get read when your prompt or the CLAUDE.md index points to them, which is exactly why the index is worth writing.
Do I need a design system before I can document design tokens?
No. Document whatever visual decisions you’ve already made, even informal ones. If you know your accent color is #E94560 and your font is Inter, that’s enough to start. A partial token file beats no token file, and you can fill in radius, shadow, and spacing values as those decisions get made.
How is this different from a system prompt?
A system prompt is a single-session instruction set. A brand context folder is persistent documentation that lives in your project and gets read across many sessions and task types. Use both: the folder for the brand baseline that should apply everywhere, the system prompt for the parameters of the task in front of you.
Can I use the same brand context folder with other AI models?
Yes, with minor adjustments. The files are plain text, so any model that accepts file context can use them. Design tokens and positioning transfer cleanly; voice files sometimes need more explicit instruction depending on the model. If you’re running the same brand across multiple models, keep one folder as the source of truth rather than maintaining a variant per model.
How long should a brand context folder take to build?
Most teams can build a functional brand context folder in 30 to 60 minutes if they already have some existing brand documentation to draw from. The voice profile and terminology file are the most time-consuming. Design tokens go quickly if you already have a design system. Start with the three core files and expand over time as you identify gaps in Claude’s output.
Key Takeaways
- A brand context folder gives Claude a persistent reference point for voice, visual identity, positioning, and ICP — so every output inherits your brand instead of defaulting to generic.
- The three core files are:
voice-profile.md,positioning.md, anddesign-tokens.md. Each should be focused (300–600 words) and concrete, with real examples. - Writing samples are more effective than voice descriptions. Include at least three excerpts from content that represents your brand at its best.
- Positioning needs a messaging hierarchy, not just a positioning statement. Without one, Claude leads with whatever claim sounds strongest in isolation.
- Calibrate before you trust it. Run one throwaway task through the folder, compare the output against your samples, and fix the file that let the miss through.
- In Claude Code, use
CLAUDE.mdin your project root to reference the brand context folder automatically across every session. - For teams that want a repeatable, automated content workflow — not just a folder that individuals remember to load — MindStudio lets you wrap this entire system into an agent any teammate can run. Start free and have a working workflow in under an hour.
