Kyvvu

Trace check

What would Kyvvu have said about your agent?

Drop a log of what your agent actually did: a Claude Code session, or the tool calls from an OpenAI or Anthropic run. You'll see, step by step, which of Kyvvu's standard policies would have blocked or flagged it, and why.

loading the engine…

The check maps each tool call to a Kyvvu behaviour from its name and arguments, then runs the same engine as the playground. The mapping is a best guess, so every step shows why it was read that way and you can correct it. The real engine, inside your agent, sees exactly what each tool does.