Agentic Coding Workshops

Claude Code Workshop: Hands-On Training for Engineering Teams

A Claude Code workshop teaches engineers to use Anthropic's agentic coding tool productively in the terminal, IDE or web: setting up project memory with CLAUDE.md, planning before coding, configuring permissions and hooks, delegating to subagents and connecting tools via MCP. The goal is a team that uses Claude Code consistently and safely on real work.

neooutput editorial team 4 min read Auf Deutsch lesen

What is Claude Code and why train on it specifically?

Claude Code is Anthropic's agentic coding tool that works in any repository, reads and edits files, runs commands and iterates on tasks, and dedicated training helps teams use its configuration options well.

Getting started with Claude Code takes minutes, but getting consistent results across a team takes deliberate setup. Features like CLAUDE.md files, permission rules, hooks, custom slash commands, subagents and MCP servers are powerful, yet many teams use only a fraction of them. A focused workshop closes that gap and helps the team agree on a shared configuration that lives in the repository.

Core topics of a Claude Code workshop

A complete workshop covers context, control, extension and workflow.

  • CLAUDE.md: what belongs in project memory, how to keep it short and current, and how nested files work in monorepos.
  • Plan first: using planning before edits for anything larger than a small fix, and reviewing the plan critically.
  • Permissions: choosing modes, allow and deny rules for commands, and why blanket approval is risky.
  • Hooks: running formatters, linters or checks automatically at defined points in the agent's workflow.
  • Custom commands and skills: packaging recurring team workflows so everyone runs them the same way.
  • Subagents: delegating focused tasks such as research, review or test writing with separate context.
  • MCP: connecting issue trackers, databases, browsers or internal tools in a controlled way.

Hands-on exercises on your codebase

Participants practice on their own repository so the outputs, such as a CLAUDE.md and shared settings, stay with the team afterward.

We typically begin by letting Claude Code explore the repository and draft a CLAUDE.md, which the team then edits down to what genuinely helps: build and test commands, architecture notes, conventions and known pitfalls. Next come real tickets: a bug fix driven by a failing test, a small feature through plan and implementation, and a refactoring guarded by the test suite. Each exercise ends with a group review of the diffs, focusing on what context or instructions would have produced a better result.

Team configuration: making results consistent

Consistency comes from committing shared configuration to the repository and keeping personal preferences separate.

Claude Code distinguishes between project-level settings that can be checked in and personal settings that stay local. In the workshop, teams decide which permissions, hooks and commands become the shared default, and document how new team members get set up. This turns individual experimentation into a team standard and makes onboarding much faster.

It also helps to agree on a small set of custom commands for recurring workflows, such as preparing a pull request description or running the full verification sequence. When these live in the repository, every engineer benefits from improvements made by one.

Safety and review practices

Claude Code can run commands and change many files, so safety habits must be built in from the first day.

We teach engineers to keep work on branches, review diffs before committing, avoid granting broad shell permissions, keep secrets out of the context, and be careful with MCP servers and external content that could carry prompt injection. Hooks and CI checks provide a second line of defense. The aim is not fear but routine: safe habits that cost seconds and prevent expensive mistakes.

Formats and follow-up

Claude Code workshops run as one- or two-day sessions, remotely or on-site, in English or German, with optional follow-up office hours.

neooutput's trainers are AI Code Engineers who use Claude Code in daily client delivery. A two-day format leaves room for advanced topics like parallel subagent workflows, headless usage in CI and custom MCP servers. Follow-up sessions a few weeks later help the team refine their configuration based on real experience. If you are considering a Claude Code rollout, a short call helps us shape the right format.

Frequently asked questions

Do participants need prior Claude Code experience?

No. We start from installation and first tasks. Participants should be comfortable with the terminal, Git and their project's build and test commands.

Which accounts or plans do participants need?

Each participant needs access to Claude Code through an account type your organization has approved. We clarify the setup with you before the workshop so no time is lost on day one.

Can the workshop cover Claude Code in the IDE as well as the terminal?

Yes. We show the terminal workflow and IDE integrations, and teams can choose what fits their habits. The core concepts are the same.

Do you also cover other tools?

On request we compare Claude Code with OpenAI Codex or Cursor, but for skill-building we recommend focusing the hands-on time on one primary tool.

All insights