Claude Code · ChatGPT Codex · OpenCode

Your team already uses AI to write code. We teach them how to review it.

Two-day hands-on workshops on your own repositories, for engineering teams across Europe. Pick the agent your team is standardising on — the review standard, the GitHub workflow and the guardrails come with all three.

Request a proposal → See the curriculum

Delivered in English and Spanish · Madrid · Barcelona · London · remote across the EU

3
Tool tracks, one review standard
2
Team transformation programs
100%
Run on your own repositories
EN / ES
Delivered natively in both
Teams we train
Fintech SaaS platforms Regulated industries Consultancies Scale-ups Public sector

Curriculum

Three tracks. Choose the one your team ships with.

All three tracks teach the same operating model — delegate, review, own — but the tooling, the failure modes and the repo conventions are genuinely different. We don't run a generic AI course with the logo swapped.

ChatGPT Codex — two days, your repository

From first run to a merged pull request the team actually trusts.

Programs

When the brief is bigger than one tool.

Two longer commitments alongside the two-day tracks: reskilling engineers into a new role, or rolling AI-assisted development out across an organisation with governance built in.

Role transformation

Forward Deployed Engineering Program

Turning senior engineers into Forward Deployed Engineers and solutions architects — the person who sits with the customer, prototypes against their stack, and ships the first working integration, using coding agents to compress that timeline from weeks to days.

Customer-facing scoping Rapid prototyping Architecture under pressure Handoff & documentation
Audience
Senior engineers, architects, pre-sales
Format
4 weeks, blended
Cohort
8–14 participants
Delivery
Onsite or remote
Request a proposal →
Team program

Team Program

The full installation: when the brief isn't "train this team on this tool" but "get the whole organisation onto AI-assisted development with a shared standard." Five progressive modules, tool-agnostic.

Spec-driven development CI/CD & IaC generation Multi-agent pipelines RAG on your own codebase EU AI Act mapping
Audience
Devs, QA, architects, tech leads, managers
Duration
6 days · 48h
Structure
Full sequence or standalone modules
Cohort
12–20 per cohort
Request a proposal →

What you leave with

Artefacts, not inspiration.

Every workshop ends with things committed to your repository. If a session produces only enthusiasm, it hasn't worked.

01

An agent rules file in your repo

AGENTS.md or CLAUDE.md written during the session: conventions, test commands, protected paths, MCP permissions and the things your codebase does that no model would guess.

02

A review checklist your team signs off on

Explicit checks for AI-generated diffs: error paths, least privilege, no secrets, dependency changes, and tests that cover the failure case, not just the happy one.

03

A working GitHub pipeline

Agent review running on real pull requests, with branch protection, required checks and a named human owner for every merge.

04

A delegation map

A written line through your backlog: what gets handed to an agent, what gets reviewed harder, and what stays human-owned regardless of deadline pressure.

05

A measurable baseline

Cycle time, review time, revert rate and escaped defects measured before the workshop, so the improvement claim after it is yours and not ours.

06

A 30-day adoption plan

Who owns the rules file, when it gets updated, how new joiners are onboarded, and a follow-up clinic where we review the diffs your team shipped without us.

GitHub integration

The workshop ends in your pull requests.

Most AI coding training stops at the terminal. Ours doesn't — the point is what happens when generated code meets your review process.

Agent review on every pull request

Codex or Claude Code running as a reviewer in CI, posting inline comments your engineers can accept, argue with or dismiss.

Rules files under version control

AGENTS.md and CLAUDE.md reviewed like code, with a CODEOWNER, so agent behaviour changes deliberately instead of drifting.

Branch protection and blast radius

Protected paths, required checks, secret scanning and least-privilege tokens — so an agent that goes wrong stays cheap.

Evidence attached to the diff

Test output, the prompt or task that produced the change, and the checks a reviewer ran — captured in the PR body, not in someone's chat history.

Quality process

How a workshop is built.

A real sequence, in this order, every time. Steps one and five are the ones most training providers skip, and they are the reason adoption sticks or dies.

01

Diagnose

We read your repository, CI and review process under NDA, and interview four to six engineers. We measure cycle time and review time before we teach anything.

2 weeks before · remote
02

Design

Exercises are written against your actual code — your framework, your test runner, your quirks. You approve the agenda before anyone travels.

1 week before · you sign off
03

Deliver

Two trainers, small cohorts, hands on keyboards. Roughly seventy per cent of the time is your team working, not us presenting.

1–2 days · 8–14 people
04

Verify

The session isn't done until the rules file, the checklist and the CI job are merged, and the team has shipped one real pull request through the new flow.

Last afternoon · merged, not slides
05

Embed

Thirty days later we review the diffs your team shipped without us, update the rules file and re-measure. That number is the honest one.

+30 days · included

Formats

Onsite, remote, or a full week.

Same curriculum, same trainers, same artefacts. The format changes with your team's geography and how deep you want to go.

Most requested

Onsite

We come to your office anywhere in Europe. Two trainers, your repository on the screen, and the conversations that only happen when everyone is in the room.

  • 1–2 days
  • 8–14 engineers per cohort
  • Spain, UK, EU-wide
  • NDA as standard
Distributed teams

Remote

Live sessions split across mornings so nobody sits through six hours on a call. Breakout rooms for exercises, shared screens for review practice.

  • 4 × half-day sessions
  • Up to 20 engineers
  • CET / GMT friendly
  • Recorded for your team only
Deep adoption

Residency week

Five days embedded with your team. We train, then we pair on real tickets and rewrite the parts of your process that the workshop exposed.

  • 5 days on site
  • Multiple cohorts
  • Process rewrite included
  • 90-day follow-up

Who it's for

Whole teams, not enthusiasts.

One engineer with a great prompt workflow is a single point of failure. We train the whole team, including the sceptics — especially the sceptics, because their objections are usually the review process talking.

BackendFrontendPlatform & DevOpsQAData & MLEngineering managersTech leadsCTOsSecurity

"The useful part wasn't the tooling. It was watching two senior engineers disagree about where the human line sits, and having to write the answer down."

Engineering lead · fintech · Madrid

Questions

The things people ask before booking.

Get started

Tell us about your team.

Send the form and you'll get a written proposal within two working days — scope, agenda, dates and price. No discovery call required first, though you can have one if you'd prefer.

Response
Written proposal in 2 working days
Languages
English and Spanish, natively
Coverage
Spain, UK and the EU · remote worldwide
Confidentiality
NDA before we look at any code
Email
hello@codexworkshops.com
Please tell us your name
Please enter a valid work email
Please add your company

No newsletter. No sales sequence. One reply from a trainer.

Request received

A trainer will read this and reply with a written proposal within two working days. If it's urgent, email hello@codexworkshops.com and say so in the subject.