HGMem is a framework designed to enhance the working memory of large language models (LLMs) using hypergraph structures. Its primary function is to tackle sense-making questions, particularly those where direct answers aren't available in the source documents. Experiments show HGMem outperforms traditional retrieval-augmented generation (RAG) methods, especially for complex queries that require deeper reasoning rather than simple text retrieval. The framework operates by dynamically creating hypergraph memory structures tailored to specific questions, which allows it to adjust to various question types and domains. To set up HGMem, users need to create a Python environment using Conda, install necessary packages, and configure their environment variables for GPU usage. The accompanying code examples guide users through building a knowledge graph and executing queries with both OpenAI's API and locally deployed models. To customize evaluations, users should format their datasets according to specific conventions outlined in the documentation. This includes organizing query files and ensuring the data structure aligns with HGMem's requirements. The article also provides a citation for related research, highlighting the ongoing academic contribution surrounding this framework.