Introduction
Six months ago, I was using ChatGPT in tabs like everyone else, but finding Anthropic Claude Code changed the way I run my business in two weeks. I had context files saved in Notion. Prompts saved in Google Docs. A growing pile of useful but isolated AI conversations that started from scratch every time I opened a new chat.
This is not a hype piece. Claude Code is a command-line tool built by Anthropic that reads your files, writes code, runs scripts, and acts on your behalf. For founders stuck running everything out of their own heads, it is the closest thing to a digital co-founder I have ever used. In this post, I will cover what it is, what it costs, who is behind it, and how I actually use it to run a business.
Key Takeaways
- Anthropic Claude Code is a command-line agent that reads your files, writes scripts, and runs work for you, not another chatbot in another tab.
- Most founders waste Claude Code by treating it like ChatGPT. The real shift is building a workspace with context, data, and reusable commands.
- Claude Code costs $100 to $200 per month on the Max plan for heavy use, less than a single hour of a contractor’s time per day.
- The product I use to run my own business sits on top of Claude Code. Context files, daily briefs, automated tasks, all in one place.
- Non-technical founders can run Anthropic Claude Code through plain English instructions. No coding required, just clarity about what you want.
- Claude Code is the engine behind real business automation: daily briefs, lead follow-up, database reactivation, reporting, all running while you sleep.
- The window to be early on Anthropic Claude Code is short. Most business owners have not heard of it. The ones who have are already ahead.
What Is Anthropic Claude Code and Why It Matters for Business Owners
Anthropic Claude Code is a command-line agent released by Anthropic, the company behind the Claude family of AI models. It sits inside your terminal, but the terminal is not really the point. What matters is what it does. It can read every file in a folder. It can write new files. It can execute scripts. It can ask you questions, wait for your answer, and pick up where it left off. It can run a 30-step task on its own and check in with you only at the moments that matter.
Compare that to ChatGPT. ChatGPT is a chat window. You type, it replies. Every conversation starts fresh. The model has no memory of yesterday. It does not know your business, your team, your numbers, or your priorities. You spend half the conversation re-explaining context just to get a half-useful answer.
Claude Code is a different category of tool. It is not a chatbot. It is an agent that operates inside your work, with full sight of your files and your data.
The Workspace Idea
The single most useful thing I did with Claude Code was stop using it like a chatbot. Instead, I built a workspace. A folder on my computer with everything Claude Code needs to act like an experienced operator inside my business. That folder includes:
- Context files that explain who I am, what Octavius does, who my team is, what my offer is, who my clients are, and how I want things written.
- Data files that pull from my CRM (Nexus), accounting software, ad accounts, and email tool every night, so the numbers are always fresh.
- Custom commands like
/prime,/email-intensive,/proposal,/draft-replythat run entire workflows on their own. - Output folders where every draft, report, and proposal lives so I can find it again.
When I open Claude Code now, it already knows what I am working on. I do not need to brief it. I type, /prime and it loads my business context, my current numbers, and the work in progress. From there, I can ask anything, and the answer comes back informed.
Why Anthropic Claude Code Changes Business Automation
For the last decade, business automation has meant tools like Zapier, Make, n8n, and platform-native workflows. Those tools are useful. I still use n8n for some things. But they share a limitation: they are workflow tools, not thinking tools. You drag boxes, set triggers, and define rules. The system does exactly what you wired it to do. The moment your process changes, you have to go back into the tool and rewire.
Claude Code is different because it can decide what to do based on what is in front of it. Hand it a messy task, and it figures out the steps. Show it your existing scripts, and it adds new ones in the same style. Ask it a question about your data, and it writes the SQL on the spot, runs it, reads the result, and answers in plain English.
The Compounding Effect
Every other tool I have used hits a ceiling. The first month is exciting. Things get faster. Then the wall comes up. The tool can do what it can do, and no more.
Claude Code compounds. Every context file I add makes every future answer better. Every script I save makes the next build faster. Every command I write turns a one-off task into a permanent capability. After three months, I had a workspace that could do things in two minutes that used to take me half a day.
This compounding is why founders who get going on Anthropic Claude Code now will be very far ahead of those who start in twelve months. The lead is not in knowing the tool. It is in the workspace you have built around it.
The Pace Principle: Layers, Not Leaps
The mistake most founders make with AI is trying to automate everything at once. They list ten things and try to build all ten at the same time. Three weeks later, nothing is finished, and the whole thing feels too hard.
The way through is layers. I built my system in five layers, each one independently useful:
- Context layer. Teach the AI who you are.
- Data layer. Connect your numbers.
- Intelligence layer. Get a daily brief.
- Automate layer. Cross tasks off permanently.
- Build layer. Use the freed time for growth.
Each layer takes a few hours of work and delivers a result. The first layer alone is worth it. By the time you are at layer four, you are looking at hours of bandwidth coming back every week.
If you want a deeper look at how I structure those layers, I have written about AI automation for founder-led businesses in more detail.
What Anthropic Claude Code Costs (And Who Actually Owns Anthropic)
Two questions come up every time I show this to a founder: how much does it cost, and is it owned by Google?
Pricing
Claude Code is included with Anthropic’s paid plans, the same plans you sign up for at claude.ai.
- Free tier. You can try Claude through the web app for free with usage limits. Free is fine for casual chat. It is not enough for serious daily use of Claude Code.
- Pro plan. Around $20 per month. Light to moderate use. Good for solo operators who are testing the waters.
- Max plan. $100 or $200 per month. This is what I run on. Heavy daily use, multiple long sessions, no real worry about hitting limits.
- API access. Pay-per-token pricing for developers building products on top of Claude.
For context, $200 per month is less than one hour of a New Zealand bookkeeper’s time per day. The system runs my entire business intelligence layer, drafts my emails, builds my proposals, writes my blog posts, and audits my numbers. Comparing it to an ops hire is honestly a bit silly. The cost is not the question.
Who Owns Anthropic
People ask if Google owns Anthropic because the funding history has been in the news. The honest answer is: Google has invested billions of dollars into Anthropic, and so has Amazon. Reports have placed Google’s stake at around 14 per cent. But Anthropic is independently controlled. The founders, Dario and Daniela Amodei, came out of OpenAI to build a company focused on AI safety. They run the show. Google is a major investor and cloud partner, not the owner.
If you want to read it from Anthropic directly, their company page covers the structure.
What Code Does Claude Use
When people ask what code Claude uses, they usually mean one of two things. Which programming languages can it write? Or, what is Claude itself built on?
On the first: Claude Code can read and write code in any major language. Python, JavaScript, TypeScript, Go, Rust, Bash, SQL, PHP, Ruby, Java. I use it mostly with Python and Bash because that is what most of my automation runs on. It also handles configuration files, markdown, JSON, YAML, and HTML without any fuss.
On the second, Claude itself is built on a transformer-based large language model trained by Anthropic. The exact architecture details are proprietary. What you need to know is that the latest models, Claude Opus 4 and Claude Sonnet 4, sit at or near the top of every coding benchmark. That is why Claude Code works as well as it does.
How I Use Anthropic Claude Code to Run My Business
The five layers I mentioned earlier are not a theory. This is how I actually use Claude Code every day.
The Context Layer
The first thing I built was a single folder of context files. Who am I? What Octavius does. My current offer. My pricing. My team. My clients. My voice. My banned words. My writing rules. All in plain markdown files.
Now, when I ask Claude Code to draft an email, it sounds like me. When I ask it to write a proposal, it follows my pricing and includes the right inclusions. When I ask it to comment on a strategy idea, it factors in the actual numbers in my business and what I am trying to do this quarter. The same query in ChatGPT gets me a generic answer. In Claude Code with full context, the answer is specific and usable on the first try.
The Data Layer
The second layer was data. I built scripts that run every night at 2 am and pull yesterday’s numbers from every system I use. Xero for accounting. Nexus for the CRM. Instantly for cold email. GA4 for the website. ClickUp for projects. Fathom for meetings. All into a single SQLite database on my machine.
Now, when I ask Claude Code “how did we do last week,” it writes a SQL query, runs it, and gives me a real answer with real numbers. No logging into six dashboards. No piecing it together from memory.
The Intelligence Layer
The third layer is the daily brief. Every morning at 7 am, I get a Telegram message that summarises everything I need to know to start the day. Yesterday’s revenue. New leads. Meeting highlights. Anything flagged as urgent. Tasks I committed to but have not done yet. It takes me five minutes to read. By 8 am, I am fully informed without having checked a single dashboard or scrolled a single inbox.
I have written more about the daily brief that runs my mornings and how I built it.
The Automate Layer
This is where the real-time recovery happens. Once context, data, and intelligence are in place, I started crossing tasks off permanently. A few examples from my own workspace:
- Drafting follow-up emails after a discovery call. Used to take 15 minutes. Now takes 30 seconds.
- Writing weekly performance reports for ad clients. Used to take 90 minutes per client. Now runs in the background and lands as a draft.
- Cleaning and structuring messy CRM data so the rest of the system can act on it cleanly.
- Drafting proposals from a discovery call. Two hours down to twenty minutes.
Each one is a recurring task that used to come back every week. Now they happen, and I review the output. The model does the work. I check it.
The Build Layer
This is the layer most founders never get to because they never build the first four. The whole point of layers one to four is that they buy you back hours every week. Layer five is what you do with that time. In my case, growth, building products, spending time with my kids, and working out properly again. It’s the kind of bandwidth I had honestly given up on a few years ago.
If you want to see how the first four layers fit into a wider operation, I have written about the practical implementation plan for AI in a small business.
The Practical Starting Point: Your First Build with Anthropic Claude Code
If you are reading this and thinking, “This sounds useful, but I would not know where to start,” here is the actual sequence I recommend.
Step 1: Install Claude Code
Sign up for a Claude Pro or Max plan at claude.ai. Install the Claude Code CLI by following Anthropic’s install instructions. It runs on Mac, Windows, and Linux. Takes about ten minutes to set up. No coding required for installation, just running a few commands they give you.
Step 2: Build Your First Context File
Open a new folder on your computer. Call it whatever your business is called. Inside that folder, create a single file called CLAUDE.md. Open it in any text editor and write down, in plain English:
- What your business does
- Who your customers are
- What your current offer is and how it is priced
- Who is on your team, and what do they do
- Your top three priorities this quarter
- Your writing voice (formal, casual, blunt, warm)
- Anything else a new ops hire would need to know in their first week
Aim for one to three pages. You can always add more later. This file alone, before you do anything else, will make every future Claude Code conversation ten times more useful.
Step 3: Run Your First Real Task
Open Claude Code in that folder. Ask it to do one real task you would normally do yourself. A draft of a follow-up email. A first pass at a blog post. A summary of a long PDF you have been putting off reading. Watch what comes back. Edit it. Tell Claude what to change. Ask it to do another version.
That first task is the moment most founders realise this is different. It is not a chatbot. It is an operator that already knows your business, working alongside you.
Step 4: Keep Going
Once the first context file works, the next moves are obvious:
- Add more context as you notice gaps
- Connect a piece of data (start with one source, like your CRM)
- Build a custom command for a task you do every week
- Add a daily brief once you have a few weeks of data flowing in
Each step delivers value before you start the next one. Layers, not leaps.
If you want a wider look at the tools I use alongside Claude Code, I have written about the AI tools I actually run my business on.
Conclusion
Anthropic Claude Code is the biggest shift in how I run my business since I started using a CRM. It is not another tool to add to your stack. It is a layer that wraps around your work and starts thinking with you. Most founders are still using AI like a chatbot. A few have realised that the real win is building a workspace where the AI knows your business and can act inside it.
The bigger picture matters here. Claude Code is the engine. The brain of your business is what you build on top of it: context, data, intelligence, automation. That combination is what gets a founder out of the operator trap and back to the work that actually grows the company.
If you want help mapping what that looks like for your operation, the AI Strategy Intensive is a two-hour 1:1 session where I look at your business with fresh eyes, map every recurring task, and show you exactly where Claude Code can be installed first for the highest return. Book it at octavius.ai/intensive. $797, fully credited against any work we do together if you take the next step.
Frequently Asked Questions
What is the Anthropic Claude Code?
Anthropic Claude Code is a command-line agent built by Anthropic, the company behind the Claude family of AI models. It runs in your terminal but acts more like a digital operator than a chatbot. It reads files, writes scripts, runs commands, and completes multi-step tasks on its own, checking in with you only when it needs your input. It is designed for serious work, not casual chat.
Can I use the Claude code for free?
You can try Claude itself through claude.ai with a free tier, but Claude Code as a working agent really needs a paid Pro or Max plan. Pro is around $20 per month and is enough for light use. Max is $100 to $200 per month, and is what most heavy users run on. There is also pay-per-token API access for developers building products. The free tier alone is not enough for daily business use.
Does Google own 14% of Anthropic?
Google has invested billions in Anthropic and is one of its major partners alongside Amazon. Reports have placed Google’s stake at around 14 per cent, though that figure shifts with new funding rounds. Importantly, Google does not own or control Anthropic. The company is independently run by its founders, Dario and Daniela Amodei. Google is a significant investor and cloud partner, not the owner.
What code does Claude use?
Claude Code can read and write code in every major language: Python, JavaScript, TypeScript, Go, Rust, Bash, SQL, PHP, Ruby, Java, and more. It also handles configuration files, markdown, JSON, YAML, and HTML. The Claude model itself is built on a transformer-based large language model trained by Anthropic. The architecture details are proprietary, but the latest models sit at or near the top of independent coding benchmarks.
How is Anthropic Claude Code different from ChatGPT?
ChatGPT is a chat window. Every conversation starts fresh, and the model knows nothing about your business unless you re-explain it each time. Claude Code is an agent that lives in a folder on your computer. It reads your files, sees your data, and remembers your context across every conversation. Instead of typing back and forth, you give it tasks, and it completes them. Different category of tool entirely.
Do I need to know how to code to use Anthropic Claude Code?
No. The name is a bit misleading. Claude Code is a command-line tool, but you talk to it in plain English. Tell it what you want, and it figures out the technical steps. I have watched non-technical operators install it and run their first real task inside an hour. The skill is not coding. It is being clear about what you want and being willing to iterate when the first answer is not quite right.
What can a small business actually do with Claude Code?
Quite a lot. The use cases I see most often are: a daily brief that summarises yesterday’s numbers, drafting follow-up emails after sales calls, writing reports automatically from CRM data, reactivating dormant leads, generating content drafts in your voice, audit-style checks across spreadsheets, and one-off scripts for messy admin tasks. The pattern is consistent. Boring, repetitive work gets crossed off permanently. Time goes back to the founder.
About Octavius
Titus Mulquiney is the founder of Octavius AI, where he builds AI brains and AI workforces for founder-led businesses stuck running everything out of their own head. Twenty years in marketing, ex-Sony product manager, ex-GM Zeal NZ. Based in Auckland, working with operators across NZ, Australia, and the US. Connect on LinkedIn.