How to Build a Brand Identity File for Claude Code: The AI Interview Method
Instead of writing your identity file from scratch, let Claude interview you. Here's how to create user.md, soul.md, and brand context files in minutes.
The Problem With Writing Your Own Brand Identity File
Most people stare at a blank user.md or CLAUDE.md for ten minutes, write three vague sentences about their “professional tone,” and call it done. Then they wonder why Claude Code keeps missing the mark.
Brand identity files for Claude Code — the context files that tell the AI who you are, how you write, and what your brand stands for — are only useful if they’re specific. But specificity is hard to produce on demand. You know your brand instinctively. You don’t necessarily know how to write it down.
The fix is simple: don’t write the file yourself. Let Claude interview you, then turn your answers into a finished identity file. This guide walks through the exact process for creating user.md, soul.md, and supporting brand context files using the AI interview method — and why it produces better results than anything you’d write from scratch.
What Brand Identity Files Actually Do in Claude Code
Claude Code is Anthropic’s agentic coding tool that runs in your terminal and can read, edit, and create files across a project. Like most AI coding assistants, it works best when it understands context — not just your codebase, but you.
That context lives in a few key files:
CLAUDE.md— Project-level instructions: coding conventions, architecture notes, preferred libraries, workflow rulesuser.md— Personal context: who you are, your role, how you prefer to communicate, what you care aboutsoul.md— Brand or creative identity: voice, values, audience, tone, what the brand stands for
Other agents start typing. Remy starts asking.
Scoping, trade-offs, edge cases — the real work. Before a line of code.
These files aren’t part of every Claude Code setup by default. Some developers use just CLAUDE.md. Others build out a full identity layer that includes personal and brand context. How much you invest depends on how often you use Claude Code for anything beyond pure syntax — content, copy, documentation, strategy.
If you’re using Claude Code to write anything that faces an audience — docs, emails, marketing copy, product descriptions — the brand identity layer pays off quickly.
Why Starting From Scratch Fails
Here’s what usually goes wrong when you try to write your own brand identity file:
You describe what you want, not what you actually do. “We use a professional but approachable tone” describes almost every brand on earth. It gives Claude nothing useful.
You forget the things that feel obvious. Your writing instincts are baked in after years of practice. You don’t think to write them down because you’d never break those rules yourself. Claude doesn’t share your instincts.
You write generically because you’re writing into a void. Without a prompt pushing you toward specifics, most people default to safe, abstract language. The result reads like a brand guidelines template with your name on it.
The interview method solves all three problems by externalizing the process. You respond. You don’t generate.
The Three Files Worth Building
Before running the interview, it helps to know what you’re building toward.
user.md — Your Personal Context File
This file tells Claude about you as a person and a professional. It covers:
- Your role and area of expertise
- How you prefer to communicate (direct vs. exploratory, formal vs. casual)
- What you already know well (so Claude doesn’t over-explain)
- What you hate in AI output (hedging, filler phrases, unsolicited caveats)
- Your working style and preferences
A good user.md is 200–400 words. Specific, not exhaustive.
soul.md — Your Brand or Creative Identity File
This file captures what your brand actually is. It’s harder to write and more valuable when it’s right. It covers:
- Brand voice and tone with real examples
- Core values (the ones that actually drive decisions, not the ones on your About page)
- Audience — who you’re writing for and what they care about
- What you never say or do
- The brand’s personality if it were a person
Expect soul.md to run 400–700 words when complete.
A Supporting Brand Context File
Optional but useful: a file that covers factual brand context — company name, product names, key terminology, common use cases, competitor landscape, and any other reference material Claude should keep handy. Call it brand-context.md or context.md. This is the easiest to write manually because it’s mostly factual.
Setting Up the Interview: The Core Prompt
Here’s the method. You open a Claude conversation — not Claude Code, just Claude — and give it this setup prompt:
I want to create a brand identity file called soul.md for use with Claude Code.
Instead of writing it myself, I'd like you to interview me.
Ask me one question at a time. Wait for my answer before asking the next question.
Cover these areas: brand voice and tone, core values, audience, what the brand
stands for, what it never does, and its personality.
After I've answered all your questions, synthesize my answers into a complete,
specific soul.md file in markdown format. Make it concrete — use my exact words
and examples where possible, not abstract summaries.
Start with your first question.Built like a system. Not vibe-coded.
Remy manages the project — every layer architected, not stitched together at the last second.
Then just answer the questions honestly. Don’t overthink it. One to four sentences per answer is enough. Claude is pulling the specific language out of your responses — you don’t need to produce polished answers.
Running the Interview: What to Expect
A full brand identity interview runs 10–20 questions. Claude will ask things like:
- “How would you describe the tone of your best piece of content?”
- “What’s something a competitor does that you’d never do?”
- “Who is your ideal reader — what do they already believe before they find you?”
- “If your brand were a person at a dinner party, how would they talk?”
- “What’s a phrase or word you’d use to describe what you do that most people in your industry wouldn’t use?”
These questions are designed to pull out specifics. Your job is to answer honestly and resist the urge to give “correct” answers. If you catch yourself writing what sounds good rather than what’s true, stop and rewrite.
The interview typically takes 15–25 minutes. When you’re done, tell Claude to generate the file.
Running the Same Process for user.md
Use a similar prompt but scoped to personal context:
I want to create a personal context file called user.md for use with Claude Code.
Interview me to gather the information. Ask about my role, expertise, communication
preferences, what I already know well, and what frustrates me in AI output.
One question at a time. When done, synthesize into a complete user.md in markdown.The user.md interview is shorter — typically 8–12 questions, 10–15 minutes.
Refining the Output
Claude’s first draft of your identity file will be good but probably not perfect. Here’s how to refine it:
Check every abstract claim against your actual work. If the file says “direct and concise,” find a piece of your real writing. Does it match? If not, change the file.
Cut anything that’s true of every brand. “We value quality” tells Claude nothing. Delete it or replace it with something specific — “we’d rather publish fewer pieces than pad out a calendar with thin content.”
Add examples. If there’s a piece of writing that perfectly captures your voice, paste a paragraph directly into the file with a note. Real examples outperform any description.
Test it. Ask Claude Code to write something — a short email, a product description, a section of documentation. Compare it to your own voice. Note what’s off and add correction notes to the file.
The file improves fastest through use, not through editing in isolation.
Where to Store Your Identity Files
Claude Code looks for CLAUDE.md in specific locations: the project root, parent directories, and your home directory (~/.claude/CLAUDE.md for global settings). Your user.md and soul.md can live in a few places:
Global home directory — Best for personal context and voice that applies across all projects. Put user.md in ~/.claude/ and reference it from your global CLAUDE.md.
Project root — Best for project-specific brand files. A content marketing project might have its own soul.md that’s different from a technical docs project.
A central /context folder — Some teams keep identity files in a dedicated folder and reference them explicitly in CLAUDE.md using import-style instructions.
Whatever system you use, the key is making sure Claude Code actually reads these files. Test this by asking Claude Code directly: “What do you know about my brand voice?” If it draws on your files, you’re set.
Scaling This Process With MindStudio
Once you’ve built identity files for yourself, the obvious next step is doing this for clients, teammates, or multiple brands — which is where manual interviews get slow.
MindStudio is a no-code platform for building AI agents and workflows. One practical application here: build an interview agent that runs this entire process automatically. The agent asks the same structured questions, collects answers, and generates a finished soul.md or user.md at the end — all without you facilitating each session manually.
With MindStudio’s visual workflow builder and access to models like Claude, you could build:
- A client onboarding agent that interviews new clients about their brand and outputs a ready-to-use identity file
- A team setup workflow that walks each team member through a personal
user.mdinterview and stores the result - A brand audit agent that reviews existing brand content and generates an identity file based on patterns in the writing
MindStudio connects to 1,000+ tools including Google Docs, Notion, Airtable, and Slack — so the generated files can automatically land wherever your team’s workflow lives. You can try MindStudio free at mindstudio.ai and build the first version of this kind of agent in under an hour.
If you’re already using AI agents to automate content workflows, a brand identity interview agent slots in naturally as a setup step before any content generation runs.
Common Mistakes to Avoid
Making the file too long. Claude Code has context limits. A 3,000-word identity file competes with your codebase for attention. Keep each file focused. If it’s getting long, split it.
Using the file as a style guide. A style guide is a reference document for humans. An identity file is active context for an AI. Write for a reader who processes instructions literally, not one who interprets nuance.
Forgetting to update it. Brands change. Voices evolve. Re-run the interview process every 6–12 months, or whenever you feel like Claude’s output is drifting from what you’d actually write.
Treating negative instructions as enough. “Never use jargon” is a start, but “use plain English — if a 10th grader wouldn’t know the word, replace it” is more useful. Positive instructions outperform prohibitions.
Not testing with real tasks. The only way to know if your identity file works is to use it. Run a real task and read the output critically. Adjust based on what’s wrong, not based on what sounds better in the abstract.
Frequently Asked Questions
What’s the difference between CLAUDE.md and soul.md?
CLAUDE.md is for project and workflow instructions — how Claude Code should behave technically. It covers things like “use TypeScript strict mode,” “run tests before committing,” or “don’t modify files in the /legacy folder.” soul.md is for brand and creative identity — voice, values, audience, personality. They serve different purposes and both are worth having.
Do I need separate files or can I put everything in CLAUDE.md?
Remy doesn't build the plumbing. It inherits it.
Other agents wire up auth, databases, models, and integrations from scratch every time you ask them to build something.
Remy ships with all of it from MindStudio — so every cycle goes into the app you actually want.
You can consolidate if you prefer simplicity. A single CLAUDE.md with sections for project rules, personal context, and brand voice will work. Separate files become more useful when you want to reuse identity context across multiple projects without copying it — keep user.md and soul.md in your home directory and project-specific rules in each project’s CLAUDE.md.
How specific should my brand voice description be?
As specific as possible. “Confident and direct” could describe thousands of brands. “We don’t hedge — we make claims clearly and stand behind them, even when the topic is complex” describes one. Aim for descriptions that would differentiate your brand from your five closest competitors.
Can this method work for technical documentation, not just marketing copy?
Yes. The interview questions look different — you’d focus more on technical accuracy standards, audience expertise level, and documentation conventions — but the principle is the same. You answer, Claude synthesizes, you refine. Building AI agents for documentation workflows follows naturally from having a solid context file in place.
How often should I update my identity files?
At minimum, once a year. More practically: update whenever you notice Claude’s output consistently missing something important, or when your brand or role significantly changes. The re-interview process is faster the second time because you’re refining, not starting from scratch.
Does the interview method work with other AI coding assistants?
The general principle works anywhere you can provide system-level context. Cursor, Copilot Chat, and similar tools all support context files or system prompts. The specific file names and locations differ, but the interview-then-synthesize method for generating that context is the same.
Key Takeaways
- Brand identity files for Claude Code only work if they’re specific — generic descriptions of “professional tone” give the AI nothing useful to work with.
- The AI interview method produces better files than writing from scratch because it externalizes the process and pulls specific language out of conversational answers.
- Build three files:
user.mdfor personal context,soul.mdfor brand identity, and optionally abrand-context.mdfor factual reference material. - The interview prompt is simple: ask Claude to interview you one question at a time, then synthesize your answers into a finished file.
- Refine by testing — run real tasks and adjust the file based on what Claude gets wrong, not what sounds better in isolation.
- For teams or agencies, MindStudio makes it practical to turn this interview process into a repeatable automated workflow.
If you’re already using Claude Code daily, better context files are the fastest way to get better output — and the interview method gets you there in under an hour. Start with soul.md, test it on a piece of real writing, and build from there.
