Background
Many development teams already know autocomplete assistants like Copilot — but that is fundamentally different from working with an agent that reads your codebase, proposes a plan and implements changes on its own. Management has often already understood why agentic software development matters. What's missing is the how at the team level: how do you set up Claude Code properly? What permissions does an agent actually need? How do you build trust without ending up in a state where nobody quite knows what the agent changed?
We built a dedicated workshop for exactly that team level — the counterpart to our decision-maker workshop: that one is about shared understanding and where AI fits the business, this one is about the workbench itself.
Format and audience
Two days, tailored to the client. On-site is preferred: a room where everyone sits together noticeably works better than a hybrid setup — hybrid is possible, but the second-best option. Group size follows the team: for larger groups we bring additional trainers so that everyone still gets individual help when something gets stuck.
The audience: software developers and software architects — anyone who codes and wants to sharpen their own skills working with agents. Research groups too, where code and publication are the same piece of work anyway. This is not a management workshop, not a slide deck for decision-makers. Anyone looking for the strategic overview belongs in our AI workshop; this one is about getting hands dirty.
A variant for research groups
For research groups, day two shifts: instead of a product repository, the group's own work takes centre stage. We show how findings get recorded for the paper as they appear, rather than being reconstructed afterwards from notes and terminal history. How your own bibliography can be pulled in as context, so an agent argues from the sources you already work with. And how all of it lands in a second brain — Notion or Obsidian, say — so that intermediate results don't get lost inside individual chat transcripts.
Sample agenda
The agenda isn't a fixed curriculum — it adapts to participants' prior knowledge and their project. Day one might look something like this:
- Install and set up Claude Code
- Build a shared understanding — our guest lecture "Agentic Software Development" at HHU Düsseldorf serves as the starting point, more inspiration than script
- Work through a sample project together
- Let agents work autonomously
- Build in issue triage
- Set up automations
- Wire in connections with n8n
Day two shifts gears: instead of practising on the sample project, we keep working together on the participants' own real project — with the actual questions, code and quirks that come with it.
What participants take away
The focus is consistently hands-on: participants work at their own machine instead of watching a demo. That's why support scales with the group — bring more people and you get more trainers, so there is still time to look over everyone's shoulder individually.
At the end of the two days there's no abstract certificate, but a shared way of working across the team: how a task gets handed to an agent, how the result gets reviewed, where your own role shifts from typing to directing and checking — and where the decision deliberately stays with the human. Plus a first working setup for automation and issue triage that keeps running on the participants' own project once the workshop is over.
Teams we have trained
We have already trained Daniel Jung Media GmbH, VIA – Analyse, Training & Therapie GbR, ordino and FAYA UG (haftungsbeschränkt) in Claude Code. All four now build their own apps. That is the yardstick we measure the workshop against: not whether everyone is impressed at the end, but whether they are still working this way weeks later.
Why this fits us
In 2026 Christian Meter gave the guest lecture "Agentic Software Development" in the "Software Development in Practice" series at Heinrich Heine University Düsseldorf. Before that he was a research associate there at the Computer Networks chair, and taught programming and software development courses himself from 2015 to 2021. The lecture is the starting point of this workshop — the workshop then goes where a lecture hall can't: to the real project, with the questions that only come up while working.