OpenAI Codex – Agent-based Software Development in Practice (AICODX)

AI - Artificial Intelligence, AI for developers

One-day practical workshop for experienced developers, architects and tech leads who want to use OpenAI Codex as an autonomous development agent within real team workflows. Hands-on sessions cover agent workflows, prompt design, and integration in IDE and terminal environments.

Course focuses on hands-on exercises with repositories, code review, refactoring and tests, plus setting security guardrails and team rules. Expect live demos of repo workflows, code review patterns and practical tips for safe agent adoption.

Location, current course term

Contact us

Custom Customized Training (date, location, content, duration)

The course:

Hide detail
  • Introduction to OpenAI Codex
    1. What OpenAI Codex is and how it works
    2. Evolution of AI tools for developers
    3. Differences between ChatGPT, Copilot, Cursor, and Codex
    4. Agent approach versus classic code generation
    5. Typical practical use cases
    6. Limits of current coding agents
    7. The developer's role in the era of AI agents
  • Basics of agent-based development
    1. What an AI agent is
    2. Differences between chatbots, workflows and agents
    3. Lifecycle of an agent task
    4. Context, planning, reasoning and actions
    5. Tool use and working with external tools
    6. Single-agent and multi-agent architectures
    7. Human-in-the-loop approaches
  • How a coding agent thinks
    1. Decomposing complex tasks
    2. Planning solution steps
    3. Working with the repository and project context
    4. Finding relevant information in code
    5. Iterative problem solving
    6. Self-checks and result validation
    7. Sources of errors and hallucinations
  • Prompt engineering for Codex
    1. How coding prompts differ from regular LLM prompts
    2. Structure of a high-quality instruction
    3. Defining goals and expected outputs
    4. Limiting task scope
    5. Specifying technical constraints
    6. Iterative prompt refinement
    7. Common mistakes when working with Codex
  • Instructions and project context
    1. Project-level instructions
    2. System instructions
    3. AGENTS.md and project rules
    4. Organizing project knowledge
    5. Versioning instructions
    6. Sharing instructions within the team
    7. Minimizing inconsistent outputs
  • OpenAI Codex in practice
    1. Overview of Codex environments and capabilities
    2. Working on an existing repository
    3. Analyzing project architecture
    4. Code navigation techniques
    5. Generating code changes
    6. Refactoring and modernizing code
    7. Generating documentation
    8. Explaining unfamiliar code
  • Practical developer workflows with Codex
    1. Requirement analysis
    2. Solution design
    3. Implementing new features
    4. Generating unit tests
    5. Refactoring existing solutions
    6. Debugging and fixing issues
    7. AI-assisted code review
    8. Preparing a Pull Request
    9. Documenting changes
  • TDD and code quality with AI
    1. Test-driven development with an agent
    2. Generating test scenarios
    3. Covering critical cases
    4. Reviewing generated code
    5. Verifying outputs
    6. Static analysis
    7. Using AI as a reviewer
  • Advanced workflows and multi-agent approach
    1. Delegating tasks among multiple agents
    2. Roles of specialized agents
    3. Architect agent
    4. Developer agent
    5. Tester agent
    6. Reviewer agent
    7. Coordinating agent workflows
    8. Practical multi-agent collaboration scenarios
  • Tool integration and MCP
    1. What the Model Context Protocol (MCP) is
    2. Connecting AI agents to external tools
    3. Integrating GitHub, Jira, Confluence and others
    4. MCP servers and their roles
    5. Practical integration examples
    6. Securing MCP access
    7. Limits and risks of external integrations
  • Security and governance
    1. Risks of agent-based development
    2. Prompt injection attacks
    3. Handling secrets and sensitive data
    4. Agent permissions
    5. Approval workflows
    6. Auditing and activity logging
    7. Security guardrails
    8. Team policies for AI use
  • Open-source models and hybrid approaches
    1. Cloud vs local models
    2. OpenAI models for development
    3. Open-source coding models
    4. Ollama and OpenAI-compatible endpoints
    5. Hybrid architectures
    6. Cost and performance trade-offs
    7. When to choose a local model
  • Comparing AI-assisted development tools
    1. OpenAI Codex
    2. GitHub Copilot
    3. Cursor
    4. Claude Code
    5. Windsurf
    6. Continue.dev
    7. Strengths and weaknesses of each tool
    8. Recommended tool combinations by project type
  • Final hands-on exercise
    1. Analyze an existing repository
    2. Plan change using Codex
    3. Implement a new feature
    4. Generate tests
    5. Perform code review
    6. Prepare a Pull Request
    7. Share experiences and best practices
  • Technical requirements
    1. Visual Studio Code
    2. Git
    3. Access to an OpenAI account with Codex
    4. Sample repository for exercises
    5. Recommended: Docker and local dev environment
  • Participant requirements
    1. Practical software development experience
    2. Familiarity with Git
    3. Basic command-line (CLI) skills
    4. Experience with at least one programming language
    5. Prior use of ChatGPT, GitHub Copilot or similar is a plus
Assumed knowledge:
Practical software development experience, familiarity with Git and basic CLI.
Schedule:
1 day (9:00 AM - 5:00 PM )
Language: