Qt's Coco MCP Server Preview is a Model Context Protocol server that exposes Coco code coverage CLI tools as structured MCP tool calls for AI coding agents. It solves the problem of AI agents lacking runtime coverage data when generating tests by providing exact per-instrumentation-point execution data. Key tools include check_coverage_summary, import_execution_report, get_coverage_overview, get_file_coverage_detail, and analyze_patch_coverage. The server uses progressive disclosure to minimize token usage and supports C++, QML, and mixed Qt projects. A workflow example shows parser.cpp coverage improving from 47.6% to 88.9% using AI-generated tests guided by the MCP server. Future plans include CRAP metrics, Python support, and QML code coverage.