Claude Projects setup in 30 minutes: the no-code version
The exact 30 minutes to set up your first Claude Project, with the voice file, the knowledge files, and a working test prompt. No coding, no terminal, all browser-based. By the end, every Claude conversation will sound like it works for you, not generic AI.
30 minutes once. By the end you have a Claude Project with your voice file, your knowledge files, and a working test prompt. Every chat inside the Project will sound like your business instead of generic AI. The single biggest unlock available to Claude Pro users in 2026, and 90% of Pro subscribers we audit haven’t set one up. Pro required ($30 AUD/month).
Why this matters
Most Pro subscribers use Claude exactly the way they used the free tier: open a chat, type a prompt, get an answer, close the chat. They’re paying $30 a month for capability they’re not touching.
Claude Projects is the capability you’re not touching. It’s a folder that holds your business’s context (voice file + sample writings + brand guidelines + FAQ). Every chat inside that folder starts with Claude already knowing all of it. You don’t re-paste your menu, your past captions, your tone rules every conversation. It just knows.
Setup is 30 minutes. The payoff lasts forever. Here’s the 30 minutes.
Minute 0-5: Open Claude.ai and create the Project
Open claude.ai in your browser. Make sure you’re signed in to your Pro account.
In the left sidebar, click Projects. If you don’t see Projects, you’re on the free tier; upgrade to Pro at claude.ai/upgrade or stop here and come back when you’ve upgraded.
Click Create Project. Name it after your business or your role. Examples:
- “Marlowe’s Cafe Brunswick”
- “Northcote Physio”
- “Bennett Plumbing”
- “Jenn’s Consulting”
Click Create. You’ll see a blank Project with two sections: Instructions and Knowledge.
Minute 5-15: Write the voice file (in Instructions)
The Instructions field is what Claude reads at the start of every chat. This is the most important text you’ll write all year.
Paste this template, then edit. Aim for 200-500 words.
WHO I AM
I run [BUSINESS NAME], a [TYPE OF BUSINESS] in [SUBURB], [CITY], Australia.
We do [WHAT YOU DO IN ONE SENTENCE]. Our customers are mostly [WHO YOUR
CUSTOMERS ARE: age, situation, vibe].
WHO I WRITE TO
Most of my writing is aimed at [PRIMARY AUDIENCE: existing customers,
prospects, both]. They know [BACKGROUND ASSUMPTIONS]. They don't know
[THINGS WE SHOULDN'T ASSUME]. They care about [WHAT MATTERS TO THEM,
in their words].
TONE
We sound [3-5 ADJECTIVES: warm, dry, slightly cheeky, plain-spoken].
We don't sound like [3-5 NEGATIVES: corporate, salesy, gushing,
American, over-friendly].
HARD RULES
- No em-dashes
- Australian English (organised, colour, centre, recognise)
- AUD currency unless I say otherwise
- DD/MM/YYYY dates
- No exclamation marks (or: only when [SPECIFIC CASE])
- Banned words: any AI-tell words I'd never use myself
VOICE SAMPLES
Three to five sentences that sound exactly like me, copied from real
past writing:
1. [paste one]
2. [paste one]
3. [paste one]
4. [paste one]
5. [paste one]
OFF-LIMITS
- [Topics, claims, or angles I never touch]
- [Competitors I don't name]
- [Confidential things]
OUTPUT DEFAULTS
- Unless I say otherwise, keep replies under [N] words
- Default format: [paragraphs / dot points / structured headers]
- If you're not sure of something, ask before generating instead
of guessing
Spend the full 10 minutes here. Bad voice file = generic output. Good voice file = Claude sounds like you wrote it.
Minute 15-25: Upload knowledge files (in Knowledge)
The Knowledge area takes documents (PDF, DOCX, TXT, MD) and links them to every chat in the Project. Upload 5-10 well-chosen files.
Recommended for most operators:
- Product / service list with descriptions, pricing in AUD, common questions
- 5-10 past customer reply emails you were proud of (paste them into one file, label each)
- 2-3 past blog posts or social captions you’d publish again (voice samples)
- Your About Us page copy (so Claude knows the business backstory)
- Your FAQ (so Claude can answer customer questions in your voice)
- A list of words you never use (separate from voice file for easy maintenance)
Don’t bulk-upload your whole Google Drive. Project performance degrades with too many files; 5-15 well-chosen beats 50 average ones.
Minute 25-28: Save the Project
Click Save. The Project is now active. You can switch between Projects from the sidebar.
Minute 28-30: Test it
Click New Chat inside the Project. Run a real writing task:
“Draft a polite reply to a customer who emailed asking about our hours / pricing / availability for next week. Use my voice.”
Read the output. Then open a fresh non-Project chat in Claude and run the same prompt without context. Compare.
The Project version should sound like you wrote it. The non-Project version sounds like generic AI. If the difference isn’t obvious, the voice file is too thin, add 2-3 more samples and 2-3 more rules.
What changes after setup
Three concrete differences:
1. Output sounds like you, every time
Before Projects: every chat starts cold, every reply is generic, you spend 2-3 minutes re-explaining your business in each conversation.
After Projects: chat starts with Claude already knowing your voice, your products, your tone rules. First reply is on-brand.
2. You stop copy-pasting context
Before Projects: “Here’s my menu, here’s my brand voice, here’s what we do, here’s an example of a good caption, now write me three Instagram captions for our winter menu.”
After Projects: “Write me three Instagram captions for our winter menu.”
The second prompt is 12 words; the first is 200+. Multiply by 50 conversations a month.
3. Output quality compounds
Most operators discover that month 1 in a Project, output is good. Month 3, output is great (because they’ve refined the voice file based on what Claude got wrong). Month 6, output is borderline indistinguishable from their own writing.
The voice file is a living document. Open it every month. Add what Claude keeps getting wrong. Add new no-no words. Refresh voice samples.
Multi-brand operators: one Project per voice
If you run multiple brands (or wear multiple hats: founder, parent, side hustle, board role), make one Project per identity. Don’t merge them.
Why: voice contamination. A merged “brand A + brand B” Project produces blended output that sounds like neither brand cleanly. Separate Projects keep the voices distinct.
We do this internally: Boring Ventures Project (parent co), DotVA Project (the VA agency), Lead Gen Empire Project, On Autopilot Project. Switching between them is one click in the sidebar.
What this doesn’t do
Be honest about limits.
- Doesn’t replace expertise. A voice file makes Claude sound like you. It doesn’t make Claude know what you know. Subject-matter expertise is still you.
- Doesn’t replace strategy. Voice = tone. Strategy = direction. Different things.
- Doesn’t auto-learn from new chats. The Project doesn’t update itself based on new conversations. You update it manually as you notice patterns.
- Doesn’t work on the free tier. Pro only.
For the things it does do, it’s the highest-use 30 minutes of AI setup most operators will run all year.
What’s next
- Build a social media powerhouse with Claude Code, Sunday-afternoon Claude Code build that pairs with your Project setup.
- The first 10 prompts every Australian SMB owner should run, copy-paste prompts to test against your Project.
- How to fine-tune AI for business voice, the deep dive on voice files (the 90-minute version of what we did here in 10).
- Claude for the not-quite-beginner, the broader intermediate guide.
Common questions
What's the difference between Projects and just having a long chat?
Can I have multiple Projects?
Is Projects on the free tier?
What if my business is just me, no separate brand?
Will Anthropic train on my Project content?
Want this built for your business?
Book a free 30-minute AI audit. We'll map your business and show you exactly which systems we'd build first. No pitch deck, no scoping fee.
Book my free AI audit