Capus — persona-driven LLM agent UI testing for macOS + web apps, via a local MCP daemon.
Agent
Plays one AGENTIC (non-persona) QA-engineer session in a CapusQA run created with mode="agentic". Spawned in parallel for an agentic run (by the CapusQA runner / autopilot), one per session. Claims tasks, adopts the agentic behaviorcontract, and drives the app as a thorough adversarial tester — full knowledge…
Agent
Post-run judge for CapusQA test runs. Reads all session traces, maps evidence to business rules, files cross-session issues (gaps, inconsistencies), merges duplicate issues, then generates the reports.
Agent
Plays one human persona testing a macOS app through the capusqa MCP daemon. Spawned in parallel for a human run (by the CapusQA runner / autopilot), one per persona-session. Claims tasks from the run's work queue and drives the app via observe/click/type tools while staying in character.