Concepts

MarkLogic MCP Server integrates MCP client AI assistants with MarkLogic Server databases through the Model Context Protocol (MCP). This section explains how the system is structured, how requests move between components, how access to data is protected, and how the MarkLogic Retrieval API finds and ranks content.

Design principles

The MarkLogic MCP solution is designed around these principles:

• Separation of concerns:
  • The MCP client handles the user interaction and agent reasoning.
  • MarkLogic MCP Server handles MCP communication, tool exposure, request routing, and authentication behavior.
  • MarkLogic Retrieval API performs search and retrieval operations in MarkLogic Server.
• MarkLogic Server security enforcement: Requests are executed with the authenticated user's identity. MarkLogic Server enforces role-based access control and document permissions when content is searched or retrieved.
• Use of MarkLogic Server capabilities: MarkLogic Retrieval API uses MarkLogic Server search indexes, vector capabilities, and relational views rather than implementing a separate search platform.

In this section

• Architecture describes the system components, the relationship between MCP tools and MarkLogic Retrieval API endpoints, request flows, and common deployment topologies.
• Security model describes authentication modes, trust boundaries, credential handling, request validation, and MarkLogic Server access control.
• Retrieval and ranking model describes full-text and vector retrieval, filtering, result merging, scoring, and topk behavior.