Agentic Coding Hackathon - Claude Code Edition

Transform your coding skills: From prompt to AI-based agentic software development process + agent harness system
➕ I’m interestedto the learning content

 

Brief description

The Agentic Coding Hackathon from Obvious Works is an intensive, practice-oriented training for AI-supported software development.

In 3 days you will learn how to work with CLI-based AI tools such as Claude Code, Codex or comparable agents, the Model Context Protocol (MCP) and a professional Agent harness system develop production-ready software - from the idea to deployment.

⏱️ Total time required from 3 days (4-5 days depending on AI level)
🏷️ from CHF 3,499 per participant
💳 Payment by credit card, Paypal or by invoice
🥇 „Agentic Coding Professional“ certificate“
👥 Cohort-based approach with a maximum of 12 participants
💬 German or English

    Includes:

    • Course materials & harness templates
    • Best Practice Guides & How-Tos
    • Access to the learning platform (LMS)

     

    🔗 Quick links

    All details...
    Learning content
    Start dates
    References
    Q&A about the training

    Discover our revised Agentic Coding Hackathon:

    Learn Claude Code, MCP and Harness for product-ready AI features in 3-5 days.

    Discover our revised Agentic Coding Hackathon:

    Learn Claude Code, MCP and Harness for product-ready AI features in 3 days.

    + note interest

    Voices from the boot camp

    💬„What used to take over 6 months, we can now do in less than a month - with higher test coverage and fewer errors.“

    - A DevAI bootcamp participant / software developer

    💬 „Coding via agents was the biggest aha moment for me.In contrast to classic prompting, the agent works iteratively -

    Code is built, checked and revised several times.“

    - A DevAI bootcamp participant

    💬 „We have reduced major code changes, which normally take more than a week, to a single day.“

    - A DevAI bootcamp participant

    💬 „The AI is now much more focused.Thanks to the set of rules we learned at the boot camp,

    she stays on task much better.“

    - A DevAI bootcamp participant

    About

    Graduates in leading DACH companies.

    Voices from the boot camp

    💬„What used to take over 6 months, we can now do in less than a month - with higher test coverage and fewer errors.“

    - A DevAI bootcamp participant / software developer

    💬 „Coding via agents was the biggest aha moment for me.In contrast to classic prompting, the agent works iteratively -

    Code is built, checked and revised several times.“

    - A DevAI bootcamp participant

    💬 „We have reduced major code changes, which normally take more than a week, to a single day.“

    - A DevAI bootcamp participant

    💬 „The AI is now much more focused.Thanks to the set of rules we learned at the boot camp,

    she stays on task much better.“

    - A DevAI bootcamp participant

    About

    Software developers trained in leading companies.

    Why this hackathon?

    The future of software development is agentic. Instead of simple prompts, we build autonomous AI systems, who plan, implement, test and iterate independently.

    This hackathon shows you how to use CLI-based AI tools such as Claude Code or Codex so that your AI no longer just suggests code - but Projects fully implemented.

    What makes this hackathon unique

    Tool agnostic: The concepts work with Claude Code, Codex, Aider and comparable CLI-based AI tools. You learn principles, not tool dependencies.

    Practice instead of theory: Over 80% of the time you work hands-on on real projects. At the end, you go home with a functioning AI system.

    Focus with enterprise depth: Developed for the requirements of European companies - with a focus on data protection, security and sustainable workflows.

    Proven: Over 400 graduates. Participants report 6× faster deployments and 90% fewer bugs after the training.

    Request course information

    „The days of us developers writing code are largely over. We can now take on a different role and perspective.“

    - Matthias Herbert, AI Transformation Expert @ Obvious Works

    Short call with Matthias?

    „What you learn in this hackathon will be the production standard tomorrow.“

    - Matthias Herbert, AI Transformation Expert @ Obvious Works

    Agentic Coding Hackathon

    * This hackathon is a concentrated form of our established 12-week DEV AI bootcamp. Contents and structure are roughly outlined here and will be adapted to the needs of the participants and the respective software development process on a case-by-case basis. 

     

    🧠 Contents & structure* from 3 days

    can look as follows.
    If 4-5 days are required depending on the AI level, the content will be adjusted.

    Agentic Foundations

    The first day lays the foundation for professional AI-supported development. You will understand the architecture behind agent-based systems and build your first agent harness.

    What you can expect:

    • Introduction to the Agent harness principle - the control layer for consistent AI agents
    • Structure of a Agent.md with rules, commands and dos/donts
    • Establishment of CLI-based AI tools (Claude Code, Codex) and MCP servers
    • Context Engineering: Token efficiency, Markdown docs, subsystem notes
    • CLI-first vs. tool calling: When which approach makes sense
    • Blast Radius Estimation: Small, scoped changes for controlled development

    Mini-Challenge: Build your first AI Harness - from prompt to functioning agent system

    Specs, TDD & RAG

    • AI-First Delivery Loop: You will learn what professional feature specs look like and how test-driven development works with AI agents.What you can expect:
      • Creation of professional Feature specs according to the WHAT/HOW/WHERE/WHY framework
      • The complete AI-First Delivery Loop:
        • Frame the Change (Blast Radius + Suggest Options)
        • Plan-by-conversation (short prompts, co-planning, green light)
        • Build with Guardrails (atomic commits, insert errors verbatim, fail-fast)
        • Stabilize (Caching, Background Workers, Comments-as-Spec)
      • Test-driven development in the AI-supported development process (Red-Green-Refactor)
      • RAG integration for data-driven agents: retrieval pipelines with vector databases
      • Identify and implement MCP integrations - so that AI agents can dynamically decide when to look where

      Mini-Challenge: TDD Speedrun - feature with complete delivery loop and automated tests

    Quality, Safeguarding & Hive Mind Hack

    • You will learn systematic review processes, debugging frameworks and, optionally, how to get started with multi-agent systems.
    • What you can expect:
      • 3-step review system: Syntax → Logic → Architecture
      • Debugging framework: Hypothesis → Evidence → Root Cause → Fix
      • Pre-commit pipelines with security scans (Gitleaks, Semgrep)
      • Terminal UX & Infrastructure: Automated rebuilds, stable terminals, minimal tooling
      • Human Factors: Avoid AI addiction, work-life balance, hype check
      • Multi-agent orchestration with Claude Flow and Hive Mind topologies
      • Coordination of agents in swarm systems with collective memory

      Optional: Safeguarding & Responsible AI - Secure by Design, Privacy by Design, risk analyses, ethical checklists

      Final challenge: Complete feature with delivery loop, review system and optional multi-agent setup

     

    ⚙️ Tech stack & tools

    • CLI-based AI tools: Claude Code, Codex, Aider and similar agents
    • Model Context Protocol (MCP): For dynamic tool integration - GitHub, databases, browsers, IDE tools
    • Editors & IDEs: VS Code, Cursor, Windsurf, IntelliJ, Visual Studio with GitHub Copilot CLI
    • RAG infrastructure: PostgreSQL with pgvector, Chroma, Pinecone
    • Frameworks: TypeScript/Next.js, Python/FastAPI
    • Security & Quality: Pre-Commit Hooks, Gitleaks, Semgrep
    • Multi-Agent: Claude Flow v3, Ralph Wiggum Extension

    🧪 Learning objectives

    • A complete Agent harness set up with rules and specs

    • CLI-based AI tools Connect with MCP and local tools

    • Configuring the MCP server, so that agents decide dynamically when to use which data sources

    • TDD in the AI-supported development process apply and generate automated tests

    • RAG-based agents build and connect with vector databases

    • The AI-First Delivery Loop (Blast Radius, Plan-by-Conversation, Guardrails, Stabilize)

    • Context Engineering for token efficiency and better AI results

    • Multi-agent systems with Hive Mind and Swarm topologies

    • Sustainably improve your development processes AI-First convert

    👥 Target group

      • CTOs and tech leads, who want to establish AI-first strategies in their teams

      • Senior developers, who want to multiply their productivity with agentic workflows

      • AI Engineers, who want to build harness-based architectures professionally

      • Development teams, who want to use Claude Code, Codex or other CLI-based AI tools productively

      • Startups and companies in the DACH region with an AI-first strategy

      Prerequisites: Solid programming skills (JavaScript, TypeScript, Python or comparable) and initial experience with LLMs (Claude, GPT, Codex or similar)

    In-house training

     

    In-house training makes sense for 6 or more participants. We are also happy to come to your company and bring our training with us. Simply make a get in touch with us.

     

    In-house training as live remote training

    Are your employees spread across several company locations and working from home? We offer you the opportunity to conduct a fully-fledged in-house training course online. Our experienced trainers teach the same material as in a face-to-face training course and are of course available to answer participants' questions.

    Interactive exercises in small groups are carried out in the same quality via breakout rooms and Miro as a collaboration platform. You also save on travel costs for the trainer.

     

    We cater to individual wishes

    Do you have special focuses or interests for your team? We are happy to cater to your individual requirements.

     

    Subsequent workshop with your team

    Would you like to apply what you have learned with your employees in your own context after the training and need a workshop as a booster? We would be happy to conduct a workshop at your company and address your individual priorities. Contact us us without obligation.

    Further links

     

    Bootcamp platform / LMS

    Github Copilot

    Tabnine AI Coding Assistant

    Windsurf Editor

    Ultimate Prompt Generator by Obvious Works

    The Blog from Obvious Works

    Take a look at our videos on the topic

    Your friendly trainers

    Matthias Herbert

    to the trainer page

    Michael Mey

    to the trainer page

    You don't wanna go alone?

    We understand, bring your mate!

    🎁 1x Voucher for you
    🎁 1x Voucher for your buddies

    25 CHF voucher each for Amazon / Digitec or Galaxus

    Arrange a hackathon info call
    {

    "I highly recommend Michael! His expertise and teaching style made learning about IREB not only manageable, but truly enjoyable. He has an incredible ability to break down complex concepts into clear, understandable components, making everyone in the session feel confident and well prepared. Michael's passion for the subject shines through in every aspect of his teaching. He is patient, approachable and always willing to answer questions or clarify doubts. His real-world examples and practical insights made the material come alive and gave me a deeper understanding of the IREB framework. Thanks to Michael's guidance, I now feel well equipped to apply what I have learned in my professional role. If you are considering a course or workshop with Obvious Works, don't hesitate - you are in the best hands."

    Yancza I. from Zurich

    Click here

    "Very likeable and good lecturer and trainer. Michael explains the content with incredible calm and composure. He engages individually with his course participants and answers questions in a practical way. Bravo. Please keep it up."

    Sven S. from Bern

    "Michael understood how to address our different employee groups and how to tailor a digital workshop to our specific needs. We were impressed by the mixture of theory and practical examples with which he addressed the tasks of our foundation. The uncomplicated, professional collaboration resulted in an interesting webinar in a pleasant atmosphere."

    Head of Human Resources at a large charitable foundation

    "The work with the participants was very interactive, not just frontal teaching. Good mix between lecture, discussion, practical exercises and gamification. The content of the course was conveyed in a great and understandable way."

    Ulrike L. from Zurich

    "The work with the participants was very interactive, not just frontal teaching. Good mix between lecture, discussion, practical exercises and gamification. The content of the course was conveyed in a great and understandable way."

    Ulrike L. from Zurich

    "Good organization of the course, understandable content. In my opinion, the trainer managed to address everyone in their own way despite the very different levels of knowledge of the participants."

    Goia K. from Basel

    "Very good documentation and examples, Michael was very competent and conveyed the rather dry material with humor - was great - thank you very much!"

    Daniel H. from Heilbronn

    "I enjoyed the course and found the combination of theory and practice very valuable. Thank you very much for 3 intensive days of knowledge transfer."

    Alfred P. from Frankfurt a. M.

    "The trainer was well prepared and gave a clear and understandable presentation and involved the participants. I particularly enjoyed the personal/professional exchange with other participants and trainers (especially with regard to practical examples and experience)."

    Jasmina D. from Zurich

    "The trainer created a pleasant atmosphere and learning environment. He responded flexibly to all participants and the hybrid setup of the course worked very well."

    Oliver K. from Munich

    References

    FAQ - Questions & Answers about Agentic Coding Hackathon

    Do I need previous knowledge?

    Yes - solid coding skills and initial experience with LLMs (Claude, GPT, Llama or similar) are recommended.

    What is agentic coding and why is it so revolutionary?

    Agentic coding means that AI models such as Claude Code not only generate code, but also plan, test and refactor it independently. Instead of prompt responses, autonomous workflows are emerging - a new era of software development in which developers collaborate with AI agents as they would with colleagues.

    What exactly will I learn in the Agentic Coding Hackathon?

    You will learn how to connect Claude Code with the Model Context Protocol (MCP), build harness systems, apply Test-Driven Development (TDD) with AI and integrate RAG (Retrieval-Augmented Generation) for data-driven agents. At the end of the hackathon, you will build a productive, autonomous AI system.

    What is an agent harness and what role does it play?

    An agent harness is the control layer around the AI agent - consisting of instructions, tools and user messages. It defines how the agent works and makes decisions. Harness-based architectures are crucial for consistency, security and reproducibility in AI-supported software development.

    How does Claude Code differ from other AI coding tools?

    Claude Code works agentically - it plans, queries, tests and optimizes independently. In contrast to tools such as Copilot or Cursor, Claude Code uses MCP servers, rule files and Agent.md rules to implement complex projects in a structured and comprehensible way.

    What is the Model Context Protocol (MCP)?

    The Model Context Protocol (MCP) is an open standard that connects AI agents with tools - e.g. GitHub, databases or browsers. It works like a „USB-C for AI integrations“. In the hackathon, you will learn how to configure your own MCP server and use it to extend Claude Code.

    How does RAG (Retrieval-Augmented Generation) work in the hackathon?

    RAG combines AI models with knowledge databases or vector memories. This allows Claude Code to make data-based decisions and provide context-relevant answers. In the hackathon, you will build a RAG pipeline with PostgreSQL or Chroma - perfect for data-driven agents and AI-first workflows.

    How does the hackathon integrate security and responsible AI?

    We embed security, data protection and ethics directly into the design phases („Secure by Design“, „Privacy by Design“). You will learn how to carry out risk analyses and privacy impact assessments - so that your AI is not only efficient, but also responsible.

    What is the „Hive Mind Hack“ with Claude Flow v3?

    In the optional part (day 3), you will build a multi-agent system with Hive Mind and Collective Memory. Claude Flow v3 enables the coordination of multiple agents in swarm topologies - a practical introduction to autonomous AI ecosystems and agentic orchestration.

    Do I need previous knowledge or can I take part as a beginner?

    You should have solid programming skills (JavaScript, Python or similar) and some experience with LLMs (Claude, GPT, Llama or similar). The hackathon is not for absolute beginners, but ideal for developers who want to apply AI in practice.

    What distinguishes the hackathon from online courses or other bootcamps?

    Online courses are often theoretical - here you implement real projects in 3 days. You will work with Claude Code, MCP and Harness systems in teams, receive direct feedback from mentors and go home with a productive AI system.

    How does the hackathon help me in my job or company?

    You will learn how to integrate AI workflows into existing development processes - from automated tests to the code review pipeline. Participants report 6× faster deployments, 90 % fewer bugs and higher code quality after the training.

    Will I receive a certificate and documents after the hackathon?

    Yes, after the hackathon you will receive the „Agentic Coding Professional“ certificate, all harness templates, MCP configs and RAG examples as downloads - plus access to our LMS for refreshers and updates.13.

    How does this differ from Claude Flow?

    Claude Flow v3 is part of the optional Hive Mind Hack (Day 3) - for participants who want to test multi-agent systems.
    Request hackathon information
    ★★★★★
    ★★★★★
    (4.7 out of 5)

    Don't buy a pig in a poke

    Talk to us beforehand!

    Message via contact form
    Book free call (15 min)
    Contact us via WhatsApp