Agentic Software Engineering

Overview

This innovative course introduces agent-based software development where AI functions as an active peer programmer. Participants learn how to establish continuous collaboration with an established IDE Extension, where the AI directly modifies, generates, tests, builds, and executes code. The course covers advanced strategies such as planning & acting phases, rules formulation, prompt techniques, context management, and extending agent capabilities through MCP servers. Participants will work in existing software projects with single page application frontends, REST backends, databases and API integrations, and look at fixing issues, debugging, implementing new features, understanding code and building a project specific agentic workflow context framework. Participants will also build their own little MCP server, to get a strong understanding of how models can be given access to tool usage. There will be an overview on how to select models, choose providers or approach internal hosting. The course focuses on a well established, open source, vendor and model provider independent AI integration in Visual Studio Code. Alternative AI focused IDE's, Plugins or Integrations will be discussed. To make the maximum use of time, we'll focus on one tool with different models, but the concepts, workflows and approaches are meant to be transferable to any tool with the same or even stronger capabilities.

What you will achieve

Upon completing this course, you will be able to integrate AI agents as active peer programmers in your development workflow. You will master techniques for effective context building, rule formulation, and agent interaction across both planning and execution phases. You'll gain practical experience working with AI agents on real-world software projects, including frontend and backend development, testing, and debugging. Additionally, you'll understand how to extend agent capabilities through custom MCP servers and develop project-specific frameworks that enhance your team's productivity. These skills will enable you to leverage AI not just as a tool but as a collaborative partner throughout the software development lifecycle.

Who it is for

The training is aimed at experienced software developers, tech leads, and architects who have basic knowledge of AI-assisted development and want to take the next step toward agent-based collaboration. The course is particularly suitable for developers who want to increase their productivity through AI agents that actively participate in the development process and can directly manipulate, test, and execute code.

Course format

The course combines theoretical concepts with intensive practical exercises and projects. Participants work on real development tasks using a strong agentic coding tool with AI capable of interacting with any part of the software project. The trainer guides the process with expert knowledge and individual support to foster optimal collaboration between developer and AI agent.

Requirements

This advanced course requires solid programming skills and experience in software development with object oriented languages like Java, TypeScript or C#. Participants should be familiar with common development environments, version control systems, and build processes, as the AI agents will interact directly with these tools.

Course content

The course will consist of the following topics and may be extended or adapted based on the audience. The examples in the course will focus on widely used programming languages, software architectures and frameworks. For in-house courses there is a selection of programming languages and frameworks which can be chosen to better fit the audience.

• Introduction to Agentic Software Development
  • Evolution from "one-shot prompting" to agentic workflow peer programmers
  • Understanding the agentic workflow paradigm
  • Overview of course environment and tools
• Fundamentals of IDE-Integrated AI Agents
  • Setting up and configuring the IDE extension
  • Understanding planning vs. acting modes
  • Effective communication patterns with AI agents
• Context Management and Rule Formulation
  • Building effective project context
  • Defining clear rules and guidelines
  • Advanced prompt engineering techniques
• Working with Existing Codebases
  • Strategies for introducing AI agents to complex projects
  • Code comprehension and documentation generation
  • Refactoring and optimization with agent assistance
• Practical Development Tasks
  • Debugging and fixing issues with AI agents
  • Implementing new features in existing applications
  • Test creation and validation
• Frontend Development with AI Agents
  • Single page application component development
  • UI/UX visual validation and improvement
  • Browser use and end to end testing
• Backend Development with AI Agents
  • REST API design and implementation
  • Agentic API testing approaches
  • Database integration and interaction
• Use and build MCP Servers
  • Understanding the Model Context Protocol
  • Implementing tool-using capabilities
  • Integrating external services and APIs
• Model Selection and Deployment
  • Comparing different AI models and capabilities
  • Cloud providers vs. self-hosting considerations
  • Performance and cost optimization
• Advanced Agent Interaction Techniques
  • Multi-step reasoning and planning
  • Handling complex requirements
  • Error recovery and iterative improvement
• Building Project-Specific "Agentic Tooling Framework"
  • Sharable rules, context and tooling
  • Agentic empowering testing setup
  • AI powered dev-containers
• Future Trends and Best Practices
  • Emerging technologies in AI-assisted development
  • Security considerations
  • Team collaboration with AI agents

The course focuses on a well established, open source, vendor and model provider independent AI integration in Visual Studio Code. Alternative AI focused IDE's, Plugins or Integrations will be discussed. To make the maximum use of time, we'll focus on one tool with different models, but the concepts, workflows and approaches are meant to be transferable to any tool with the same or even stronger capabilities.