Proceedings of the International scientific and practical conference ―Science in the Era of Globalization‖ (May 22-24, 2026) / Publisher website: www.naukainfo.com. - Zurich, Switzerland, 2026. - 353 p.

122 Fig. 1. MCP Server Architecture Thus, the system implements a clear end-to-end chain from a controlled client request to the delivery of structured results, with explicit separation of responsibilities between orchestration, source adapters, transformation, and persistence (Fig. 1. MCP Server Architecture) 4. Implementation of the MCP Server for OVDP Data Collection A separate data collection adapter is implemented for each bank or broker, with a unified MCP layer on top. The overall process logic is consistent (run scraper, normalize data, retrieve result), while the specific acquisition method depends on the source type: public API, authenticated API, or a dynamic web page with JavaScript rendering. Before any web scraping is performed, the target website‘s ‗robots.txt‗ file is checked to ensure that crawling is permitted and to comply with access policies on the website. 4.1. PrivatBank (Bank) A headless browser scraping approach is implemented via Puppeteer for the government bonds catalog page. Data is extracted from DOM structures representing bond rows and includes bond name, ISIN, maturity date, coupon/yield