What's sharp and new:
- Full CLI support - run/create/update tests using the command line
- MCP Server [!] - let Claude/Cursor run your prompt tests (yes, AI testing AI)
- Tag your test cases - because "test_final_v2_REAL" wasn't working
- Token-based API auth - for when you want to automate everything
- Security improvements - (locked down the fish tank)
- Optimized MCP payloads - fewer tokens, more savings
Your AI agent can now test your prompts, creating test cases directly from your Claude Code call, run tests, and more.