Send the same prompt twice. Left pane: through Seracade with routing on. Right pane: directly to your LLM provider, no routing. Side-by-side responses and costs. The visible answer to "does the routed call match the quality of the baseline."
Supports OpenAI and OpenRouter keys for the direct pane (browser-direct CORS posture). Anthropic, Google, and xAI key shapes need a server proxy to test directly; for those keys, the right pane will fall back to a Seracade call with routing pinned to your default model.