Workshops: Agentic Coding for Roles & Industries

Agentic Coding for Medical Device Software: IEC 62304 & MDR Workshop

Medical device software is developed under rules that make every requirement, change and test traceable, and that does not change when an AI agent writes part of the code. This workshop shows MedTech and health software teams how to use coding agents such as Claude Code, GitHub Copilot or Cursor within IEC 62304, the EU MDR and an ISO 13485 quality management system, without creating gaps in the technical documentation.

Who is this workshop for?

  • Software teams at medical device manufacturers, including software as a medical device
  • Quality management and regulatory affairs (QM/RA)
  • Software architects and verification and validation engineers
  • Health IT and digital health providers handling patient data
  • Development partners working for MedTech companies

What you will take away

  • Position coding agents within your IEC 62304 software lifecycle and your QMS processes.
  • Keep traceability from requirements through risk controls to code and tests when agents contribute.
  • Define a risk-based approach to validating AI tools used in development.
  • Adapt review, verification and documentation depth to the software safety class.
  • Keep patient and health data out of prompts and use synthetic test data instead.

Agenda

  1. Module 1

    Regulatory context

    • IEC 62304: lifecycle processes and software safety classes A, B and C
    • EU MDR, ISO 13485 and ISO 14971: where AI-assisted development touches them
    • EU AI Act: AI tools in development versus AI inside the device
  2. Module 2

    Agents in the lifecycle

    • What agents can support: requirement drafts, unit tests, documentation, refactoring
    • Where human judgement is mandatory: architecture, risk analysis, release
    • SOUP and third-party libraries suggested by agents
  3. Module 3

    Traceability with agents

    • Linking requirements, risk controls, code and tests in your ALM tools
    • Agents draft trace matrices and spot gaps, humans approve
    • Exercise: implement a requirement with full trace evidence
  4. Module 4

    Tool validation and records

    • Risk-based validation of software tools
    • Configuration, versioning and logging of agent setups
    • Review records that hold up in a notified body audit
  5. Module 5

    Data protection and security

    • Health data as a special category under GDPR
    • Synthetic test data and anonymisation
    • Cybersecurity of connected devices and agent-written code

What changes and what does not

IEC 62304 does not prescribe how code is typed. It expects defined processes, planned verification, controlled changes and documentation proportionate to the safety class. Coding agents fit into that frame if their output is verified, reviewed and recorded like any other contribution.

The risks are specific, though: agents produce large changes quickly, may introduce new third-party components and can write tests that pass without proving anything. We show how to catch this within your existing processes, from review checklists to SOUP handling.

Validating AI tools in development

ISO 13485 expects software used in quality-relevant processes to be validated in proportion to its risk. What that means for a coding agent depends on how its output is controlled. We work out a documented rationale with your QM/RA team; the regulatory assessment remains theirs.

  • Describe the tool's intended use and limits
  • Name the controls that catch agent errors
  • Record configuration and versions
  • Re-assess when the tool or model changes

No product code, no patient data

Exercises run on a sandbox repository modelled on a medical device project, so neither product code nor patient data leaves your organisation. Remote or on-site, in German or English, with neooutput AI Code Engineers and AI Solutions Architects. Behind neooutput stands jocapps, with a management system certified to ISO 9001 and ISO 27001.

Frequently asked questions

Can AI-written code be used in class C software at all?

IEC 62304 does not exclude tools, but class C requires the highest rigour in design, verification and documentation. In practice: narrow tasks, intensive review, complete traceability and a well-documented tool rationale.

Do notified bodies accept AI-assisted development?

Notified bodies assess whether your processes and evidence meet the requirements, not which editor you used. What matters is that AI use is defined in your procedures and the records are complete. We give no regulatory guarantees.

Is this relevant for health IT that is not a medical device?

Yes. Hospital IT, health apps and health data platforms face similar questions on data protection, security and traceability, even when they are not regulated under the MDR. We adapt the regulatory part to your product.

Workshop

Get in touch

Want coding agents in your MedTech team without gaps in your technical documentation?

Tell us roughly how many people will join and pick dates that work – we will get back to you within 24 hours with a proposal for content, format and schedule.

Prefer to talk first?

15 minutes, free and without obligation – with our AI Trainers and AI Code Engineers.

Book a call for free (opens in new tab) +49 761 519 751 50 (24/7)

Request a training / workshop

Workshop: Agentic Coding for Medical Device Software: IEC 62304 & MDR Workshop · 1 day

Dates that work for you (optional)

Suggestions about 2 weeks ahead – pick as many as you like

Reply usually within 24 hours · jocapps is ISO 27001 certified · Privacy