GhostQA

Gives Claude the ability to run AI-driven web app tests through SpecterQA's vision-based testing engine. Instead of brittle CSS selectors, it sends screenshots to vision models that decide what to click or fill, just like a human would. You define personas and journeys in YAML, and the engine walks through your app autonomously. Claude can kick off test runs, interpret findings reports, and adjust persona configurations or journey steps based on failures. Useful when you want Claude to help debug UX issues by actually seeing your interface, or when iterating on test scenarios where natural language goals matter more than hardcoded scripts. Cost budgets and evidence collection built in, so you get screenshots and structured failure data without runaway API bills.