Let an AI agent work in the codebase — without losing control
- Plans, tasks and safe changes
- Agents, hooks, MCP and tests
- How to avoid costly mistakes
What we typically buildA new feature in your own codebase — planned, tested and ready for review.
Training · Kevin Holm
Learn to build with AI — from someone who does it every day
I train developers, teams and companies in Claude Code, Cursor, AI agents, automation and everything from frontend to backend and APIs. No slides about the future — we build real solutions, ideally in your own project, so you can carry on by yourselves afterwards.
- Online or on-site
- In your own project
- Fixed price after a short call
$ claude
> Send a confirmation email when a booking is created
● Reading the project … 14 files
● Plan: route → email template → test
✎ routes/booking.js +38
✎ mail/confirmation.html +22
✎ test/booking.test.js +41
✓ 12 tests passed
Ready for review — nothing is pushed without your approval.
Claude CodeCursorAI agentsAutomationMCPn8nAPIsBackendFrontendImplementation
3+years with AI as a core working tool — every single day
6+years in IT with code, servers, integrations and operations
100+AI tools tested and used in real projects
20+solutions built for companies across different industries
What you'll learn
From the first prompt to solutions in production
Pick one topic, or combine several into a course that fits your level and your goals.
Code faster in the editor with AI that knows your project
- Rules, context and good prompts
- Refactoring and debugging
- A shared workflow for the team
What we typically buildA shared rules file and workflow, so the whole team gets the same good answers.
Agents that solve real tasks in your systems — not just chat
- Tools, memory and control
- Human approval where it matters
- Testing, logging and security
What we typically buildAn agent that reads enquiries and drafts replies for a human to approve.
Remove the manual work: emails, data, follow-ups and reports
- n8n, Make and custom workflows
- Triggers, queues and error handling
- AI as a step in the flow
What we typically buildA flow from email to CRM to reply — like the example further down.
Get your systems talking to each other — securely and reliably
- REST, webhooks and keys
- CRM, email, payments and data
- When something fails: retries and alerts
What we typically buildAn integration between two of your systems with retries and alerts.
Servers, databases and logic that can handle real users
- Node, databases and authentication
- Deployment, operations and backups
- Security from day one
What we typically buildA small API with a database and login, deployed on a server.
Beautiful, fast interfaces built together with AI
- Modern web and components
- From sketch to finished page
- Accessibility and mobile
What we typically buildA fast, mobile-friendly page or app screen — from sketch to finished.
Bring AI into the business so it gets used — and keeps working
- Where does AI deliver the biggest gains?
- Data, roles and rules
- From pilot to everyday operation
What we typically buildA prioritised list of your best AI opportunities — and the first pilot up and running.
What it looks like
An automation you can build after one day
An email comes in. The AI understands what the customer wants, creates the case in your system, replies to the customer and notifies the team — in seconds, without anyone touching the keyboard.
That's the kind of thing we build together: small, safe flows that save hours every week and that you can adjust yourselves afterwards.
workflow · customer servicerunning
01
New email from customer“Can you make videos of our 3 new properties by Friday?”
waiting02
AI understands the enquiryRequest: 3 videos · deadline: Friday · priority: high
waiting03
Case created in CRMCustomer found · 3 tasks queued
waiting04
Customer gets a reply“We're on it — you'll have the videos on Thursday.”
waiting05
Team gets notifiedShort summary in your chat
waitingTime from email to reply: 0.0 sManual work: 0 min
How we can work together
Four ways to learn
Every course is tailored to your level, and we prefer to work on your own tasks.
1:1 coaching
You and me, online with a shared screen. We go from where you are today to you building independently.
Best for: developers, founders and specialists
Team workshop
Half a day or a full day at your offices or online. Shared tools, shared rules and a workflow everyone can use.
Best for: development teams and departments
Build day
We build one solution to completion in a day — an agent, an automation or an integration — and you learn along the way.
Best for: a concrete idea that needs to go live
Regular sparring
A fixed slot every week or month where we solve whatever is giving you trouble and take the next step together.
Best for: you, if you build every day
Would you rather learn to use AI day to day without coding? See online AI courses for businesses.
Three clicks
Find the course that suits you
Answer three questions — and get a concrete suggestion to bring to our call.
The process
How it works
- 01
Short call
15–20 minutes about where you are and what you want to be able to do. You get a fixed-price quote.
- 02
Plan
We choose topics and a real task, ideally from your own project.
- 03
We build
Hands-on from the first hour. You're on the keyboard; I show the way and catch the mistakes.
- 04
You carry on yourself
You get notes, examples and the setup, so you can keep building without me.
What you take home
Not just knowledge — something that works
- The tools set up on your own machine
- A solution you've built and finished yourself
- Rules and prompts tailored to your project
- Notes and examples to build on
- A checklist for using AI safely
- A plan for your next steps
Why learn from me
Training from practice, not from YouTube
I build every day myself
What I teach, I use in my own systems and in solutions for clients — so you learn what works now, not what worked last year.
Over 100 tools tested
I've tried most of the market. You save the time and money on the tools that don't deliver and get the ones that do.
The whole way — not just prompts
Frontend, backend, APIs, servers and operations. You learn to get the solution live and keep it running.
Honest about what AI can't do
You learn where the AI needs to be kept on a short leash: security, data, testing and the points where a human must approve.
Questions
What people usually ask
No. We start where you are. Non-technical people learn to steer the AI tools safely and build small solutions; developers learn to work significantly faster without compromising on quality.
Both. 1:1 sessions and sparring usually take place online with a shared screen. Team workshops can be held at your offices or online.
Yes, ideally. The fastest learning happens in your own codebase or your own workflows. If you don't have a project yet, we'll build one together.
A computer and access to the tools we'll be using — e.g. Claude Code or Cursor. Before we start, you get a short list so nothing eats into the day itself.
It depends on the format and scope. After a short, no-obligation call you get a fixed-price quote — with no hidden hours.