Debugg AI MCP Server

A natural-language task and a target URL drive Debugg AI's interactive browser checks. check_app_in_browser handles navigation and interaction, reporting pass or fail with screenshots; localhost targets are tunneled automatically. For route-level smoke checks, probe_page collects rendered screenshots, page metadata, console errors, and network summaries without an LLM or scenario assertions. trigger_crawl instead explores pages and navigation flows to populate the project's knowledge graph, supplying app context rather than test verdicts.

Projects and environments organize the testing context, including login credentials whose passwords are not returned. The test_suite and test_case tools let clients define cases, launch suites asynchronously, and retrieve individual outcomes; creating a case does not start a run. With executions, clients can inspect status, errors, and artifacts, including network traces and console logs from browser sessions. Those artifact links may arrive after the agent finishes, so clients can poll the execution record. Destructive deletions require confirmation, while read-only MCP resources expose project, environment, and execution details for browsing.