Two days of Agentic Engineering: the workshop for your team
On site or remote, your team works on its own topics with Claude Code. Day one produces a reviewable spec, day two a working result. The workshop is run by Tobias Boehm, who builds software this way every day.

Sound familiar?
You bought AI licenses for your team, but you are not shipping any faster. Instead, bugs and technical debt keep piling up.
Licenses in place, speed missing
Coding agents are in use, but the promised productivity never shows up. The generated code doesn't fit the architecture, and technical debt grows faster than progress.
AI without guardrails
Without clear architecture rules, every AI-assisted change becomes a risk. New features take longer, and your team spends more time on workarounds than on real progress.
Everyone uses AI differently
One tinkers with prompts, another copies blindly. Without a shared workflow and clear standards, quality varies from developer to developer, and code reviews cost twice as much.
Nobody owns the code
Without fixed review steps, generated code goes straight into the repository. Your users find the bugs first, and nobody knows which line came from where.
Sound familiar?
Who is the workshop for?
Developers
You already use AI, but everyone in their own way. After the workshop you work with a shared workflow: spec first, then small tasks, then review.
Ops and DevOps
You want to hand over routine work to agents without giving away access and secrets. Permissions, sandboxing, and guardrails are therefore a fixed part of the workshop.
Product owners
With agents, the spec matters more than the code. You learn to write requirements so that an agent can implement them and your team can review the result.
Team leads and CTOs
You bought licenses and want to see them pay off. After two days, your team has working results and a workflow it can carry on by itself.
The method works with any stack. Mobile, backend, or web: your team works on its own topics.
Recognize your team?
How the two days work
Preparation
Beforehand, a short survey shows us where your team gets stuck. That is where the tasks for the two days come from. You provide the licenses, we provide the workshop repository with test data.
Day 1: From idea to reviewable spec
Where you stand
How does your team use AI today? From chat in the browser to agents with their own skills.
What changes in development work
Code is produced quickly. Planning and reviewing become the bottleneck, and that is exactly where the workshop starts.
Context engineering
What an agent needs to know (CLAUDE.md and AGENTS.md, skills, subagents, MCP) and how to keep the context clean.
A review frame for the agent
Feedback loops, quality gates, and reviews (harness engineering) that keep generated code under control.
Security
Permissions, sandboxing, protecting secrets, and dealing with prompt injection.
Spec-driven development in groups
Each group picks a first small topic, lets the agent interview them about it, writes the spec, reviews it, and breaks it down into tasks.
Day 2: From plan to working result
Building with the agent
Claude Code turns the spec into a plan and a task list and works through the tasks one by one. The group checks every result against the spec.
Open Q&A
Everything that comes up while building is put on the table and resolved together.
Capturing the workflow as a skill
Each group captures its workflow as a skill so it doesn't depend on a single person.
Presentations
Each group demos its result live.
Does this workflow fit your team?
Why us?
Working on your topics
Your team spends two days on its own topics. At the end, there is something that runs.
From daily practice
Tobias Boehm builds with agents every day himself, most recently an entire product following exactly this workflow. In the workshop he shows what has worked and what hasn't.
Already proven
A software team at a mid-sized company has already gone through the workshop, with development, operations, and product ownership at one table.
Security from the start
Permissions, sandboxing, and handling sensitive data get their own block in the workshop.
Let's find out in 15 minutes whether the workshop fits your team. It costs nothing and commits you to nothing.
What your team gets
We tailor every workshop to your team.
- A short survey beforehand, so your team spends the two days on its real topics
- Two workshop days with Claude Code, at your office or remote
- A prepared workshop repository with test data
- Reference material: prompts plus cheat sheets on context engineering and security
- A skill pack for spec-driven development that your team keeps using in its own repositories
- For each group: a reviewed spec, a task list, a working result, and a skill
We give you the price after the initial call.
After the workshop
If your team wants to keep going, we help with that too: we set up skills and guardrails in your own repository, guide the rollout to the whole team, or support you on an ongoing basis. The AppBaumeister collective supports us here, with experience in backend, cloud, and web. We put together a separate quote for this.
A taste of the workshop
With the spec cheat sheet, you give your coding agent better tasks even before the workshop: six questions, four prompts and patterns for testable acceptance criteria, summarized from the spec part of the workshop. Free by email.
Get the spec cheat sheet