Teaching with AI language models calls for more than chat windows: it also calls for access to the programming interface — and for that, a key of their own for every student. LLMduckt provides both: teaching staff choose the models, the TeachCenter hands the API keys out to the course.
In programming courses, the interface itself is the subject matter: students write code that calls a language model. For that, LLMduckt gives each of them a key of their own for an application programming interface (API) compatible with the OpenAI API — so existing libraries and code examples keep working. For working in the browser there is Academic AI, the chat service for TU Graz staff; the two complement each other.
Teaching staff decide which models their course may use — and with that, where the computation happens. The choice includes models on TU Graz servers as well as models at external providers; the current list is published under available models (external site).
Setting it up for a course
TeachCenter support does the activation; for teaching staff it comes down to three steps:
- From the available models (external site), pick the ones the students should use.
- Contact TeachCenter (external site) support, naming the course and the chosen models.
- Once activated, a link appears in the course through which students retrieve their key and sample code.
Credentials and API keys are confidential: they must not be passed on or published — not even inadvertently in a public repository. The details are set out in the terms of use (external site).
Service launch in autumn 2026
LLMduckt is available as a service of the Educational Technology unit from autumn 2026. From then on, teaching staff can have their courses activated through TeachCenter (external site) support.