CodeHelp is an Automated Teaching Assistant
for Coding and Computer Science.

For Instructors

CodeHelp helps your students without writing code for them. It is made specifically to guide and explain but not provide solution code.

  • Accessible Support — It is always available, whenever your students are working, with immediate responses.
  • Tailored Responses — You can provide context to tailor responses to your class or even individual assignments, matching your students' needs and exactly what they are working on.
  • Valuable Insights — Quickly see where your students are having trouble and what they're asking.
  • Low Cost — Each student query costs roughly US$0.01 (see costs below).
  • Simple Integration — CodeHelp easily fits into any existing course as a complement to existing support systems.

CodeHelp is shared with everyone in your class with a simple link. There are two options:

Request an LMS Connection to connect from an LMS like Canvas, Moodle, etc.
  • Everyone will sign in automatically (no separate login) via a link from your course page.
  • Takes some time to set up, and may require support from your LMS administrator.
Sign in using Google, GitHub, or Microsoft and manually create a class from your profile page.
  • You will share an access link, and everyone will sign in using a Google, GitHub, or Microsoft account.
  • Can be set up in less than ten minutes.

The documentation provides more details.

Examples:

< >


              

              

Costs

CodeHelp itself does not take payment, but the OpenAI large language models it uses are not free. We will ask you to provide an OpenAI API key to be used for your students' queries.

Costs are low: OpenAI will charge you roughly US$0.01 for each query made with the default GPT-4o model. If your students use CodeHelp regularly and average 50 queries each over a semester (higher than the average we've observed), your total costs would be roughly $0.50 per student.

(The GPT-4o-mini model is also available, and it is much less expensive though also less accurate and helpful. The same scenario above would cost about $0.02 per student if using GPT-4o-mini.)

References

[1] CodeHelp: Using Large Language Models with Guardrails for Scalable Support in Programming Classes.
Mark Liffiton, Brad Sheese, Jaromir Savelka, and Paul Denny. 2023. In Proceedings of the 23rd Koli Calling International Conference on Computing Education Research (Koli Calling '23). DOI: 10.1145/3631802.3631830

[2] Patterns of Student Help-Seeking When Using a Large Language Model-Powered Programming Assistant.
Brad Sheese, Mark Liffiton, Jaromir Savelka, and Paul Denny. 2024. In Proceedings of the 26th Australasian Computing Education Conference (ACE '24). DOI: 10.1145/3636243.3636249