Show HN: CATT 0.2.1: Test agents/services against a Muse/Instinct-alike, in CI

A new open-source tool called CATT (Consumer Agent Testing Tool) helps developers test their services against proprietary assistants like Muse and Instinct. CATT solves the problem of testing against these agents, which are challenging due to their memory and account statefulness. The tool records full interactions, including tool calls and responses, to ensure reproducibility and reliability. It simulates approvals for service actions, tracks outcomes, and allows for optional MCP (Muse or Instinct profile) configuration. CATT's assumptions about assistant behaviors are labeled as 'documented', 'unofficial', 'assumed', or 'unknown' for developer review. The tool is designed for testing in local builds or CI, with or without LLM backing, and is actively maintained with updates as new insights about assistant behaviors emerge.