MCP Test Architect
The QA Hub Test Architect is a Model Context Protocol (MCP) server that lets AI coding assistants — such as Claude Code or Claude Desktop — autonomously generate, manage, and execute E2E test suites for your application.
Instead of manually writing test cases, you point the Architect at your codebase and it:
- Scans your file tree and discovers routes, features, and API endpoints
- Generates a structured test plan with BDD scenarios
- Writes Playwright TypeScript spec files grounded in real DOM locators
- Executes the tests and ingests results into QA Hub
- Optionally generates k6 load test scripts
How it relates to QA Hub
The MCP Architect is a client of QA Hub's public API. It uses an API token (qh_*) to:
- Push generated test cases to your project
- Submit test plans for QA Manager review
- Ingest Playwright execution results
All data flows through your QA Hub instance — the Architect itself holds no state.
Available tools
The MCP exposes 16 tools to the AI agent:
| Tool | Description |
|---|---|
analyze_architecture | Scan codebase → structured feature/route map |
generate_plan | Generate BDD test plan from architecture |
generate_specs | Write Playwright spec files from plan + locators |
generate_k6_script | Generate k6 load test script |
setup_playwright | Install and configure Playwright in the target project |
crawl_app | Browse the running app to collect DOM locators |
save_auth_session | Save authenticated browser state for test fixtures |
execute_playwright | Run Playwright specs and capture results |
sync | Push test cases to QA Hub library |
write_plan | Save test plan locally as plan.md |
write_specs | Write spec files to disk |
load_tickets | Fetch tickets from the configured connector |
select_test_strategy | Choose testing strategy (regression, smoke, etc.) |
enrich_plan | Refine an existing test plan |
run_full_cycle | Execute the complete generate → test → ingest pipeline |
bootstrap | Initialize the Architect in a target project |