Steps 1 to 4 — getting to the flashing prompt

The first four steps have one job: get you from “I read a web page” to “I have my own AI guide running on my own machine, ready to build.” They’re deliberately small — the shoe and the string — because each one has to be light enough to succeed on its own before it can pull the next into reach. That’s why it’s built this way: no big leap, no money spent before you know it fits.

Why these four, in this order

Getting Claude Code onto your machine

Claude Code is a program you run in your computer’s terminal (its text command window). Unlike the chat on this website — which lives in the cloud and can only talk — Claude Code runs on your machine, so it can read and change files there, with your say-so. Installing it is a short, guided process whatever you’re on — Windows, Mac, or Linux — with Nick walking you through the exact steps for your system. Nothing here you have to puzzle out alone.

Two AIs, working together — even across two machines

From here on you have two kinds of Claude helping, and they make a team:

• Claude in Chrome is a browser add-on: Claude can see and act inside your web browser — click buttons, fill forms, read pages. Perfect for the sign-ups and dashboards on the steps ahead. While it is optional, next steps are less easy without it.
• Claude Code is the terminal tool on your machine: it does the install, file, and server work the browser can’t.

They don’t even have to be on the same computer. Claude in Chrome might run on the laptop you browse with, while Claude Code runs on the machine you’re preparing for Beaverdam. Between them they guide you through Steps 5, 6 and beyond — one handling the clicking-side, the other the machine-side — while you watch and approve.

Who does what

At the flashing prompt? On to Step 5 — git and your first sign-ups ›

Step 5 — git, GitHub, and your first sign-ups

Relax. Working with Claude Code at that flashing prompt is really just messaging — like texting someone. You type what you want in plain words; it answers, asks, and does the work. The black window looks intimidating, but the conversation inside it is ordinary.

Picture messaging someone who was born and raised — then left all alone — on the International Space Station. They know everything about technology, and everything about the world… but nothing about living. That’s your guide: it brings the technical know-how; you bring the common sense and the real-world context about your business. Between you, it works.

First, “git” — the tool, not the company

git (lower-case) is a tool for coordinating changes to a body of work with tens of thousands of people tinkering with it, independently, all at the same time. It keeps a perfect history, lets everyone work without treading on each other’s toes, and merges in the good changes together cleanly. Beaverdam leans on git for three things: a single source of truth, a long institutional memory, and reviewed change — nothing slips in unseen.

The ways Beaverdam uses git

Where git lives — GitHub, and the alternatives

While git is the tool; you are still going to need a place online to keep and share the repositories with your staff. Several services do that job: GitHub, GitLab, Gitea (which you can even host yourself) and many others. Beaverdam uses GitHub by default — and it’s no fly-by-night startup: it hosts most of the world’s largest free and open-source projects, and since 2018 it has been a division of Microsoft. About as durable as software homes get. GitLab and Gitea are open alternatives if you’d rather; it’s a real choice, and Nick can talk you through it.

Signing up, and your SSH keys

You create a free account (a couple of minutes). Then, so your machine can have a secure, private channel into GitHub, you generate an SSH key — SSH is “Secure Shell,” and the key is a matched pair: a public half you hand to GitHub, and a private half that never leaves your control.

Guard that private half like a house key. We’ll want you to keep a copy somewhere safe — on a thumb drive tucked away, or even printed as a QR code and kept in a firesafe. Lose it and you simply make a new one, but it’s yours to protect. Nick handles the technical side of setting it up on your system.

Bringing Beaverdam home — and meeting Nick

With the account and key in place, you clone the Beaverdam project — git’s word for downloading your own working copy. Then comes a small but pivotal move: you stop Claude Code where it’s running now, and restart it inside the Beaverdam project folder. From that moment Claude Code is loaded with everything Beaverdam — and that is when it truly becomes Nick, the keeper of your own copy, rather than general-purpose Claude.

Why this is in your long-term interest

Because the system’s code, history, and reasoning all live in one open, portable place, the business is never trapped. If the person helping you changes — a year from now, or ten — the next one opens the same public history and picks up exactly where things stood, with nothing locked in someone’s memory. That is ownership and continuity made concrete: the precise opposite of the single-developer trap.

Who does what

Claude never sees or types your password, your 2FA codes, or the private half of your key — it can’t, and it shouldn’t. The account is yours, the repositories are yours, and because everything is in open, portable formats, you can take it elsewhere at any time.

Step 6 — Your control centre, and the crew to run it

This is where the building really starts — and where you meet the crew who do it with you. So, first, the introductions.

Meet the crew

Beaverdam comes with a small cast of helpers. Most of them live inside your Beaverdam — your own copy, shaped to your business (the technical word for that copy is a fork). Two others look after the shared project that every copy grows from — the original trunk from which all other forks branch off. We call those two meta-personas: they don’t work for any one business, they speak for the whole community.

In your workshop — the helpers inside your own Beaverdam

Tending the whole community — the two meta-personas

Harmonizing! (Important)

To do great work for you your staff needs to understand and gain inspiration from your vision and your mission. With AI workers it is actually much more important. AIs will happily work towards what they imagine you goal is. If they get it wrong, it will all go wrong, and you will have a hard time figuring out why.

Nick has instructions to talk you through defining your specific Mission and Vision; basically, "what you are setting out do" and "what the world will look like when you have done it". Those are the boundary lines and goal posts your AI helpers will work within and towards.

Where your control centre lives

Step 6 stands up saconsole — your systems-administration console, the control centre everything else runs from. The first job is deciding where it lives, and that’s a serious conversation. You and Nick weigh up the machines you have, what you’re comfortable running, and what it needs. Nick has done this many times and will patiently walk you through the options to find the most suitable site.

Once you’ve picked a home, Nick helps you put the plumbing in place — each piece explained as you go:

• A domain name — your own web address (like your-business.com), which you buy and own.
• A Cloudflare account — the service that points your domain at the right machine and shields it from hostile traffic.
• A wildcard SSL certificate — one padlock that secures your address and every sub-address under it (so shop.your-business.com, books.your-business.com and the rest are all encrypted at once).
• WireGuard & DNS — WireGuard is a Virtual Private Network (VPN), private encrypted pathways between your machines; DNS (the Domain Name System) is the address book that matches names to the right machine.

This is Paco’s territory. Your machines can move around — a virtual machine on a laptop today, a rented server abroad tomorrow — and Paco keeps every name pointing at the right place even after things have shifted, so you never chase a broken address.

What saconsole runs

Two things live on saconsole. One is the live map (coming up in Step 7). The other is Grafana — your monitoring: it turns the raw numbers your machines produce into readable dashboards, so the health of everything is one glance away, and it warns you before a small problem becomes an outage.

Break something on purpose

Paco and Nick set up triggers in Grafana for anything suspicious going on. Grafana and its data sources provide them with what they need in order to analyze what's happening. You should never have to get involved. However, you'll find a set of several dozen deliberate "break something" tests. You can run them at random and see how Paco and Nick handle it.

Who does what

Step 7 — The live map: seeing all your machines

With saconsole running, you get the live map — a single picture (drawn by a tool called Cytoscape) of every machine you have, colour-coded, clickable and draggable.

It’s laid out in four quadrants, so you always know whether you’re looking at devices for off-line (alpha) testing, on-line (beta) testing or approved for real live use.

The four zones of the map

Why production lives in more than one place

For the live system, Beaverdam can spread you across several providers and countries — for example an account with Contabo in the USA and another with Prometeus in Europe. A main copy (the master) is continuously replicating all data to a standby copy (the slave), so if one fails the other carries on or takes over completely.

The point of using two far-apart homes is that they share nothing that could fail at once: separate electricity, separate geography (a tsunami, tornado, or volcano that hits one can’t hit both), and separate undersea data cables (which matters in an age of deliberate cable-cutting). It’s the digital version of not keeping all your records in one building.

Those are example hosting services, not endorsements — Nick helps you choose, weighing price and fit: iwStack · Contabo · Prometeus · Apache CloudStack (the software behind pay-by-the-minute clouds).

You’ll know the moment anything happens

Put Telegram (a free messaging app) on your phone and Beaverdam sends you instant notifications of anything important — the routine good news and the rare bad news alike. For example:

• “Backup xyz recorded and validated in VM dev_003.” — a quiet all-clear.
• “Replication to Slave_002 suffered a delay of 1.73 seconds.” — a heads-up worth a glance.

None of this is improvised. The whole arrangement — the four quadrants, the replication, the multi-country spread — is worked out and very well understood by Nick and Paco; you get the expertise and the peace of mind, not the homework.

Who does what

Step 8 — actually working with ERPNext

This is the big one — and the most rewarding. By now your system is built, served, and watched over. Step 8 is the long, at-your-own-pace work of moving your business into it and learning to fly it. There’s no rush and no finish line you have to sprint to.

You grow into it, one piece at a time

You don’t swallow ERPNext whole. You start with one corner you already understand — say invoicing, or stock — get it working, gain confidence, then add the next. Experience and features grow together. The picture below hints at the breadth on offer; no business uses it all at once, and you only ever turn on what earns its place.

ERPNext also connects to the other tools a business already runs — so the pieces talk to each other instead of living in separate silos.

You plan it together — in plain words

You don’t do any of this alone, and you don’t learn ERPNext’s internals. You plan it conversationally with Nick, who calls in Wyatt when a job goes deep into the business system and Paco when it touches the network. You describe your business the way you’d explain it to a sensible colleague; they translate it into the system.

And every change is independently checked by Steph — the second-pair-of-eyes who writes the “is this really done and is this really done correctly?” test the worker never sees in advance. Quality is watched continuously, not just assumed. You stay in control of what changes and why; the crew handles the how.

A whole community has been here before you

ERPNext is used by countless businesses, with a large, friendly public community at discuss.frappe.io. When a question is genuinely new, that’s where it gets worked out in the open — and the answers accumulate for everyone. (Here’s a small example: a worked answer one of us contributed — the kind of durable, public know-how the forum is full of.)

Improvements flow back — for everyone

When your Nick (with Wyatt or Paco) works out an improvement worth sharing, they prepare it as a pull request — a tidy, reviewable proposal. Roger, keeper of the shared trunk every Beaverdam grows from, reviews it and commits it upstream, so the next business benefits too. Your fork stays yours; the good ideas lift the whole community.

Who does what

Community — the people around Beaverdam

These last spaces aren’t part of your Beaverdam — they belong to the whole community. They’re run for everyone, you’re welcome to use them, and Jono (the community builder from the crew) tends them. You just show up and join in — nothing to host, nothing to run.

The forum — durable questions and answers

The community runs a Discourse forum: an open-source discussion site built for searchable, long-lived questions and answers. Ask once, and the answer becomes part of a growing public library — so nobody has to start from scratch next time. It’s the human counterpart to the durable memory git gives the code.

The chat — quick, live talk

For real-time back-and-forth there’s a community chat. The exact tool is not yet chosen — deliberately, because the community wants an open, self-hostable one it owns rather than rents. The candidates being weighed:

Why not just Discord? It’s convenient, but closed — the community wouldn’t own its own conversations. So the choice is being made carefully rather than by default.

Announcements — on your phone, via Telegram

The community also broadcasts on Telegram — the same free app the live map uses for your system’s alerts, here used for community news. Keep an eye out for messages like:

“Hey folks — join the live streaming tutorial on YouTube, 2026/07/12 at 18:30 EST.”

Two different Telegram streams, then: one is your own system reporting on itself; this one is the community keeping you in the loop.

Who does what