Technologies / Artificial Intelligence

Codex and GitHub Copilot CLI - AI Agents in a Development Team

A four-day course on working with coding agents - Codex CLI, Desktop and Cloud, and GitHub Copilot CLI. Agent configuration through AGENTS.md, MCP tooling, Skills, sub-agents, agent teams working in parallel, code review and GitHub Actions integration. Participants build a working application, directing agents through a TDD cycle.

A four-day course on working with coding agents - Codex CLI, Desktop and Cloud, and GitHub Copilot CLI.

Codex and GitHub Copilot CLI - AI Agents in a Development Team is a four-day, intensive EITT training program, delivered on-site or remotely — depending on participant preferences. The program is aimed at experienced practitioners looking to deepen their skills — classes focus on advanced scenarios and case studies from business practice.

Topics covered include:

Day 1 - Foundations and project start
Day 2 - Tooling, sub-agents and building features
Day 3 - Agent teams, quality and production
Day 4 - GitHub Copilot CLI and working with two tools

Participants will leave the training with practical knowledge and tools for immediate application in their daily work.

Benefits

Configure Codex CLI, Desktop and Cloud as well as GitHub Copilot CLI, and choose the right mode for the task
Describe agent behaviour in an AGENTS.md file, including nested rules for individual directories
Connect MCP servers and build your own Skills and commands instead of repeating the same instructions
Build a team of sub-agents with divided roles and delegate work to them with quality control
Direct an agent through a TDD cycle in which tests are written before production code
Run an agent in non-interactive mode inside GitHub Actions with safe permission boundaries
Assess the cost of agent work and reduce it through model choice and context management

Who is this training for?

Developers and software engineers working daily with Git and Node.js

Tech leads and architects defining a team standard for working with agents

DevOps engineers wiring agents into CI/CD pipelines

Teams already using AI assistants that want to move to agent-based work

Prerequisites

Practical knowledge of Git and working in a terminal
Programming experience in TypeScript or JavaScript
Node.js installed and a GitHub account with access to GitHub Copilot
Basic familiarity with unit testing

Training program

AI in software development in 2026 - agents rather than assistants, benefits and risks (hallucination, over-reliance, context erosion, cost)
Environment check - Codex CLI, Desktop app, IDE, Node.js, Git; installing and signing in to GitHub Copilot CLI, required plan and organisation policy
Codex working modes - interactive and one-off commands, multiple sessions in Desktop, remote environments in the cloud
Model selection - quality, speed and cost; working with local and alternative models
Context management - the context window, compaction, context erosion
Permissions and security modes - sandboxing, isolation, tool approval mechanisms
From prompt engineering to context engineering - AGENTS.md as agent instructions, nested rules for different directories
AGENTS.md on the GitHub side - copilot-instructions.md and path-scoped instructions
Guiding project - an AI chat on the Vercel AI SDK, preparing a PRD and architecture decision records

Built-in agent capabilities - file work, shell, web search, browser tooling
MCP and Skills - how they differ, when to use which, and how to combine them
Configuring MCP servers - Context7 for current documentation, Playwright for browser automation, the GitHub server
Skills - installing existing ones, creating your own from session history, adding scripts and custom commands
Sub-agent architecture - inspecting threads, comparing an orchestrator model with direct interaction
Practice - defining project specialists (backend, frontend, quality) and delegating their first tasks
TDD with an agent - the red, green, refactor cycle; tests before production code
Building the application - a streaming endpoint, chat interface, code review and end-to-end tests produced by the agent

From sub-agents to teams - orchestrating multiple agents, environment isolation, conflict management
Task plan matrix - dependencies, phases and briefing agents working in parallel
Practice - three agents at once, merging results and resolving conflicts
Testing and quality - agent self-assessment, generating end-to-end tests, manual testing through Playwright MCP
Code review with an agent - security, quality, best practices; the built-in reviewer in Copilot CLI and pull request review
Agent limitations and the need for human review
Codex in the cloud - environment configuration, secrets, networking, GitHub integration
Non-interactive mode in a pipeline, CI/CD basics and application deployment
Cost and optimisation - tokens, model choice, caching

Copilot CLI from scratch - installation, sign-in, plans and organisation policy
Model selection, automatic model selection, bring-your-own and local models
Sessions - resuming, undoing changes, compacting context during long work
Instruction layers - AGENTS.md, copilot-instructions.md, path-scoped and personal instructions
Custom agents in agent.md files - role definition, tool scope, project and user level
Built-in agents, sub-agents with their own context window, planning mode and autonomous work
Hooks as quality gates and plugins for distributing configuration across a team
Native GitHub integration - issues, pull requests and running workflows from the terminal
Delegating a task to a cloud agent and receiving the result as a draft pull request
Copilot CLI in GitHub Actions - permissions, tokens, safe automation boundaries
A tool selection map and a team working standard

Delivery Methods

Online

Convenience of participating from anywhere
Interactive live sessions with trainer
Materials available for 30 days
No travel costs

On-site

Direct contact with trainer and group
Intensive hands-on workshops
Networking with other participants
Full focus on learning

Training Topics

Frequently asked questions

What is the format and duration of the "Codex and GitHub Copilot CLI - AI Agents in a Development Team" training?

The training lasts 4 days and is delivered both on-site and remotely (online) — depending on the group's preferences. The program includes lectures, demonstrations, and practical exercises, ensuring a comprehensive approach to the subject matter.

What are the prerequisites for participants?

The training is at the advanced level — it is aimed at experienced practitioners looking to deepen their skills. We expect participants to have: Practical knowledge of Git and working in a terminal, programming experience in typescript or javascript, node.js installed and a github account with access to github copilot, basic familiarity with unit testing. Detailed requirements can be found in the 'Prerequisites' section on this page.

Does the training include practical elements?

Yes, the program combines theory with practice. Participants work on case studies, take part in group exercises and practical workshops. Each exercise concludes with a debrief and feedback from the instructor.

Will I receive a certificate after the training?

Yes, each participant receives a personalized certificate of completion issued by EITT, confirming participation in the program and the competencies acquired.

What is the cost of the training and how can I register?

The training costs PLN 5 500 net per person. To register for the training or to receive a group offer, contact us by phone or through the form on our website. We also offer closed training sessions tailored to the needs of a specific organization.