Book chapter · Jagiellonian University Press, 2026
AI Agents and Model Context Protocol in Academic Library Repositories
A literature review with implementation scenarios for Polish universities. It describes what AI agents could take on in a repository (cataloguing, semantic search, recommendations, support for open science) and what has to be secured: hallucinations, user privacy, algorithmic bias, costs and new staff skills. The chapter is in Polish.
Four safeguards against hallucinations
- 1Repository data first, answer second (RAG).
- 2A source for every piece of information.
- 3Admitting uncertainty instead of guessing.
- 4A librarian approves the metadata.
- use generative AI tools46.4%
- have not adopted them yet42.9%
- report interest in the topic39.3%
Survey of 65 libraries (Jaworska and Rzeska), cited in the chapter.