Skip to content
Winter Sale from $619, until 31st December 2026 Book now
Agentic coding workshops iOS + Android

Agentic Coding Workshop: What a Good One Teaches and How to Choose

An agentic coding workshop teaches developers to hand real tasks to coding agents such as Claude Code, Codex, Cursor or GitHub Copilot and get reviewable, tested changes back. A good one runs on your own codebase, covers task framing, project memory, verification loops, code review and safe permissions, and ends with a workflow the whole team shares – not a list of prompting tricks.

8 min read

Key takeaways

  • Agentic coding is a different skill from autocomplete: the developer specifies, steers and verifies while the agent reads the repository, runs commands and iterates.
  • The most valuable workshop time is spent on your own code, with your build, your tests and your review rules – not on a toy demo project.
  • Pick a tool-specific workshop once the team has standardised on one agent; pick a tool-neutral one while you are still deciding.
  • Judge a workshop by what exists afterwards: a maintained memory file, agreed review rules, safe permission settings and merged pull requests.

What an agentic coding workshop is – and what it is not

A coding agent is more than a chat window next to your editor. It reads the repository, edits several files at once, runs the build and the tests, reads the errors and tries again. Tools like Claude Code, OpenAI Codex, Cursor's Agent and GitHub Copilot's agent mode all work this way, with different surfaces and defaults. If you are new to the idea, what is agentic coding explains the basics.

That shift changes the developer's job. Less typing, more specifying what should happen, deciding how much the agent may do on its own and checking what comes back. An agentic coding workshop trains exactly that. It is not:

  • a prompt-engineering course. Clever wording matters far less than context, task size and feedback loops.
  • a vendor demo. Watching someone else build a to-do app in ten minutes teaches very little about your monorepo with a slow build.
  • a one-off inspiration talk. The habits only stick if people practise them on real tickets and agree on team rules.

The best sessions feel like pair programming with an experienced colleague who happens to have used these agents on production code for a long time.

The six skills a workshop should build

Whatever the tool, the same six skills separate teams that get value from agents from teams that give up after two weeks.

1. Sizing and framing tasks

Agents do well with a task that has a clear goal, a known place in the code and a way to check the result. "Add pagination to the orders endpoint, keep the response format, add tests" works. "Improve performance" does not. Participants practise splitting a real backlog item into agent-sized pieces.

2. Durable project context

Every session starts from zero unless the repository tells the agent how things work. Memory and instruction files – CLAUDE.md, AGENTS.md, Cursor rules, Copilot instructions – hold build commands, conventions and no-go zones. Teams leave with a first version written and reviewed together; our guide to CLAUDE.md and AGENTS.md shows a structure.

3. Plan first, then execute

For anything larger than a small fix, asking the agent for a plan, correcting it and only then letting it write code saves hours of rework. Teams that want to go further with written specs can follow up with a spec-driven development workshop.

4. Closing the feedback loop

An agent is only as good as the signals it gets. Fast unit tests, a reliable build command, a linter and type checker turn "looks plausible" into "verified". Part of the workshop is often making these commands agent-friendly.

5. Reviewing agent output

Agent pull requests tend to be large, confident and occasionally wrong in subtle ways. Participants learn what to look for and how to keep diffs reviewable. The AI code review workshop goes deeper.

6. Permissions, secrets and tool access

Which commands may run without asking? Can the agent reach the network, production data, your cloud console? How are MCP servers vetted? These settings decide whether agentic coding is safe in your company – see the secure agentic coding workshop for the security-first version.

Talk to an AI Code Engineer about your agentic coding workshop.

In a free call we look at your team, your codebase and your goals and suggest a workshop format that fits – remote or on-site, in English or German.

Workshop formats compared

There is no single right length. The format depends on how far the team already is and how much change you want to make.

FormatTypical lengthBest forWhat exists afterwards
Introduction sessionHalf a dayLeads and teams deciding whether and how to adopt agentsShared vocabulary, a live demo on your repository, a list of open decisions
Hands-on workshopOne dayDevelopment teams starting with one toolA reviewed memory file, configured permissions, first agent-made pull requests
Workshop with follow-upTwo days plus a check-in after a few weeksTeams making agents part of the daily workflowTeam rules for review and task sizing, agent steps in CI, a retrospective on real usage
One-to-one coachingSeveral short sessionsTech leads or individuals with very specific codeA personal workflow tuned to one codebase
Internal hackathonOne to two daysTeams that already know the basicsFeatures shipped with agents, lessons captured for everyone

For a minute-by-minute example, see our sample workshop agenda. Coaching and hackathons have their own pages: 1:1 coaching and the hackathon format. Whether to meet in a room or on a video call is covered in remote vs. on-site.

Tool-neutral or tool-specific?

If your company has already bought licences for one agent, train on that tool. People learn faster when every exercise happens in the interface they will use on Monday. We run dedicated sessions for Claude Code, OpenAI Codex, Cursor and GitHub Copilot.

If you are still choosing, a tool-neutral workshop lets the team try two or three agents on the same task and compare results on your code rather than on benchmarks. The core skills transfer: a good AGENTS.md helps Codex, Copilot and Cursor alike, and the habit of planning before coding works everywhere.

The audience matters as much as the tool. Product managers, non-developers and mobile teams have different needs, which is why we also describe workshops for product managers, iOS developers and React Native and Flutter teams.

Seven questions to ask a workshop provider

  1. Do we work on our own code? A prepared sample repository is fine for the first hour. After that, real gains come from your build, your tests and your quirks.
  2. Who runs it? Trainers should ship software with agents themselves, not only teach it.
  3. How is our code protected? Ask where code goes, which tool settings are used, whether an NDA is signed and how secrets are kept out of agent sessions.
  4. What do participants need beforehand? Licences, admin approvals and installed tools should be sorted before day one, or the first two hours disappear into setup.
  5. How much is hands-on? Aim for at least two thirds of the time with hands on keyboards.
  6. What happens afterwards? A follow-up call or a short second session catches the problems that only appear in daily use.
  7. Language and location? Teams in Germany, Austria and Switzerland often want German-language sessions and clarity on works council topics – see workshops in Germany.

What drives the price, and how to judge whether it was worth it, is covered in what an agentic coding workshop costs.

How to tell whether the workshop worked

Avoid measuring success by enthusiasm on the day. Look at what changed four to six weeks later:

  • Is the memory or instruction file still maintained, or has nobody touched it since the workshop?
  • How many pull requests started with an agent, and how many of those were merged without being rewritten by hand?
  • Did review time go up or down? Agent pull requests that are too big to review are a sign that task sizing has slipped.
  • Have there been incidents – a leaked key, a deleted branch, an unreviewed dependency – that point to permission settings?
  • Do people share what worked, for example in a channel or a short weekly demo?

A short retrospective with these questions is the most useful follow-up we know.

How our workshops run

Our trainers are AI Code Engineers who use coding agents every day to build and release apps for the App Store and Google Play. Workshops run remote or on-site, in English or German, and as far as your policies allow, on your own codebase. Code handling follows the ISO 27001-aligned practices we use for client projects. Part of our workshop programme is also offered under the neooutput brand.

Because we ship mobile apps ourselves, mobile teams get the parts most trainings skip: signing, CI builds and store review. If your goal is a released app rather than a trained team, our iOS submission service takes that part off your hands. Either way, book a free call and we suggest a format that fits.

Before you book an agentic coding workshop

  • You know which agent (or which two candidates) the team will use.
  • Licences, admin approvals and installs are planned before the first session.
  • Legal or security has said which repositories may be used with coding agents.
  • The build and test commands run reliably from the command line.
  • You have picked two or three real backlog items to work on during the workshop.
  • A follow-up date is set to review how the team is using agents a few weeks later.

Frequently asked questions

Do participants need experience with AI coding tools?

No. Most groups are mixed: a few people use agents daily, others have only tried autocomplete. A good trainer uses the experienced participants as pair partners and gives the beginners a guided first task.

Can the workshop use our proprietary code?

Usually yes, if your company allows the chosen tool on that code. We clarify the tool settings, data handling and an NDA beforehand. If a repository is off limits, we prepare a realistic sample project that mirrors your stack.

How many people can take part?

For hands-on work, six to twelve developers per trainer works well. Larger groups are possible with a second trainer or by splitting into tracks, for example backend and mobile.

Is one day enough?

One day is enough to set up a team and build the basic habits. Teams that want agents in their daily workflow and CI usually add a second day or a follow-up session a few weeks later.

Free, 15 minutes, no obligation

Plan your agentic coding workshop with us.

Our AI Code Engineers run hands-on workshops on Claude Code, OpenAI Codex, Cursor and AI app builders – and ship apps to the App Store and Google Play every week. Tell us about your team.

AI tools, store policies and requirements change quickly – always check the current documentation of the tools and stores you use. This article is independent advice and not affiliated with or endorsed by Apple, Google, Anthropic, OpenAI or any other company or product mentioned; all trademarks belong to their owners.

Workshop for your team?

Talk to an AI Code Engineer – free, 15 minutes, in English or German.

Book a call for free

Chat and hotline are open 24/7. Answers from our senior AI Code Engineers (real humans) usually take closer to 72 hours than 1 hour.