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.

121 3. MCP Server Architecture for Ukrainian Government Bond Data Acquisition The system architecture is built as a sequential pipeline for controlled data collection and preparation. At the entry point is the MCP client, which formulates a request to perform a data collection or result-reading operation from file and sends it to the MCP server via the standard MCP interface. In this model, the MCP server acts as an orchestrator: it does not directly perform source parsing, but validates the client‘s intent, selects an allowed scenario, and delegates the call to the Tool API. The Tool API then serves as an application contract layer between the control layer and the specific business logic, translating an abstract command into a call to the corresponding scraping or data retrieval module. Control is then passed to the specific modules, where specific adapters are implemented for each bank or broker. These modules interact directly with the external environment: one layer connects to external sources (HTTP APIs, web pages, dynamic JavaScript-rendered content) from which raw data are extracted, while another layer feeds into the transformation layer, where these data are cleaned, normalized, converted into a consistent structure, and serialized into a tabular format. Within the transformation layer, the system transitions from source-specific representations to a unified dataset suitable for reuse and cross-provider comparison. The transformed result is then passed to the persistence layer, where the data is stored as persistence artifacts as CSV files. The persistence layer allows repeated access without re-querying external sources again and again. The final step is the delivery of data from the persistence layer to the consumer layer: the consumer (via an MCP read tool) receives already prepared content in textual form for further analysis, agent-based processing, or integration into subsequent pipeline stages.