Qt Group has released a preview of the Coco MCP Server, which wraps Coco's code coverage CLI tools in a Model Context Protocol interface for use with AI coding agents. The server exposes structured tools — check_coverage_summary, import_execution_report, get_coverage_overview, get_file_coverage_detail, and analyze_patch_coverage — that give AI agents exact per-line coverage data from real test runs. This enables feedback-loop workflows where agents identify uncovered lines, generate targeted tests, rebuild, and verify improvement. A demo on a C++ parser sample shows decision coverage jumping from 45% to 80.1% globally. The server supports all MCP-compatible hosts including GitHub Copilot and Claude Code, and is designed for C++, QML, and mixed Qt projects.
Table of contents
The Context Problem in Test Case GenerationNew to code coverage?Introducing the Coco MCP ServerTools exposed by the Coco MCP serverWorkflow examplesWhat’s Next?Need More Information?Sort: