Large Language Models (LLMs) applications in the financial domain offer enhanced efficiency in generating insights and improving accessibility of information to the public. Common applications include sentiment analysis, financial report summarization, and financial database querying. However, despite their potential, LLMs still face limitations in this domain, such as limited up-to-date financial knowledge and insufficient mathematical skills. To address these gaps, recent research has explored agentic frameworks that integrate LLMs with specialized tools—employing Retrieval-Augmented Generation (RAG) for up-to-date knowledge and calculator modules for mathematical skills. Expanding the scope of financial tasks in which agents can be applied, this work investigates the Fundamental Stock Analysis task. Specifically, we examine whether an agentic LLM can extract balance sheet and income statement data from stock press releases and compute a number of key fundamental indicators. We evaluated two LLMs, one open and one proprietary, in three prompting strategies and three agentic frameworks in 19 listed companies in the Energy Sector of the Brazilian Stock Market. Our results highlight the feasibility of automating fundamental analysis with agents, while highlighting the strengths and limitations of current LLM-based agentic approaches. The code and resources are openly available at https://github.com/AIDA-BR/fundamentalist-agent .