dukehoff

Workshops for development teams: working productively with AI agents

Your team learns to set up coding agents such as GitHub Copilot or Claude Code properly and to use them effectively and cost-efficiently every day, on your own code and with your tools. Each workshop is put together for your team from proven modules, at a fixed price.

What your team takes away

  • A working setup instead of slides: project instructions, skills, hooks and tool connections (MCP) end up in your team's repository.
  • A way of working: slicing tasks, working from specifications, reviewing agent changes, using tests as feedback.
  • Clear limits: where agents help, where they don't, and what may leave your company.
  • Costs under control: rules for model choice, context and tools, and metrics that show what an agent costs per task.

Tailored to your team

You tell us your tools and frameworks, for example GitHub Copilot, Claude Code or Cursor, Maven or Gradle, Spring or Jakarta EE. We put the modules together from that. Your team practises on its own repository or on a cleaned-up copy, not on toy examples.

The modules

Basics
How coding agents work; tools, operating models and data protection.
Setup
Project instructions (CLAUDE.md, AGENTS.md, Copilot instructions), skills, hooks, MCP servers, permissions and security.
Way of working
Spec-driven development, slicing tasks, reviewing agent changes, tests as feedback, several agents with roles.
Team and process
Team playbook, keeping costs in check (model choice, context, budgets), metrics instead of gut feeling, rollout in the company.
For Java teams
Refactorings with IDE tools, migrations with agents, debugging with agents.

Keeping costs in check

Coding agents are billed by usage, through tokens or premium requests. What a task costs depends heavily on how the agent is set up: how much context it carries along, which model it uses, and whether it has to search and try or has targeted tools.

In our public evaluation on Apache Commons Lang, renaming an overloaded method with IDE tools cost 72% less, and finding all callers of one overload cost 91% less (Claude Code with Sonnet 5, three sessions each, cost re-priced at Copilot rates). On everyday tasks such as a small feature, a bug fix or a new test, on the other hand, we saw no difference. That is why in the workshop we measure on your own tasks which setup pays off.

See the evaluation →

Three packages at a fixed price

Per group of up to 12 people, on site or remote. Preparation on your team's code is included.

  • Team Start

    €4,900

    Workshop
    1 day
    Afterwards
    2 weeks of questions by email
    Result
    Project instructions and a review checklist

  • Team Setup

    €9,900

    Workshop
    2 days
    Afterwards
    Follow-up session after 4–6 weeks
    Result
    In addition: skills, hooks, MCP, cost rules and a team playbook

  • Agents in the development process

    €15,900

    Workshop
    3 days
    Afterwards
    Two follow-up sessions
    Result
    In addition: specification template, test analysis, complete playbook

All prices net, plus VAT and travel expenses. Additional group: €3,400 per day. Office hours: €750 per month. Your own skills and tool connections at a fixed price by arrangement.

Web seminars for large groups

A series of web seminars of 2 hours each for large groups, independent of specific tools: the starting point for whole departments, before individual teams go into the workshop. With a Q&A round.

€1,900 per session

The series of four sessions €6,900. Up to 100 participants, net, plus VAT.

  • How coding agents work, and where they fail
  • Spec-driven development: from the requirement through the plan to the code
  • Assuring quality: review, tests and responsibility for AI-generated code
  • Introducing agents and keeping costs in check: roles, security, cost, measurement

How it works

  1. First call, 30 minutes, free of charge: goals, team, tools, constraints.
  2. Questionnaire and choice of modules: you tell us your tools and frameworks, we put the modules together and send a fixed-price offer.
  3. Preparation on your repository or a cleaned-up copy, under NDA if you wish. Exercises come from real tickets.
  4. Workshop: short theory, lots of practice on your own code.
  5. Follow-up: what works, what gets adjusted.

Who runs the workshop

Anton Epple, Managing Director of Dukehoff GmbH: Java Champion, author and speaker at international Java conferences such as JavaOne, Devoxx and JavaLand. Over decades he has designed and taught tailored trainings built from modules. He works with coding agents every day himself, with Claude Code and with GitHub Copilot, and builds tools for them, among them the open-source project Dukehoff MCP for Java.

Which workshop fits your team?

In the first call we clarify goals, tools and package, without obligation and in 30 minutes.

Book a first call