DICOM MCP server enables AI assistants to query, read, and move data on DICOM servers (PACS, VNA, etc.). It provides tools to perform metadata searches across patients, studies, series, and instances, and to read DICOM reports encapsulated as PDFs by extracting their text content for downstream processing. The server can send DICOM images to other destinations for AI workflows such as segmentation or classification, using C-MOVE to transfer either specific series or entire studies to configured DICOM nodes. Configuration is defined in a YAML file listing DICOM nodes and AE titles, with sample Orthanc-compatible setups to facilitate local testing. Core tooling includes query_patients, query_studies, query_series, and query_instances for metadata queries; extract_pdf_text_from_dicom for reading PDF reports; move_series and move_study for data routing; and utilities like list_dicom_nodes, switch_dicom_node, verify_connection, and get_attribute_presets to manage connections and metadata detail levels. The project supports local testing with Orthanc and emphasizes privacy, noting that DICOM-MCP is not intended for clinical use with live patient data and should be used with locally hosted open-weight LLMs.