AutoExacto

Based on the current evidence, AutoExacto looks worth watching, but the safest conclusion is that it is a clearly launched routing feature with adoption attention, not yet a broadly proven best-in-class solution. The strongest evidence comes from OpenRouter’s own posts: it continuously runs GPQA and TAU-Bench, uses the results by default for tool-call routing, and reportedly re-evaluates providers every 5 minutes across throughput and tool-call tests. X reposts and view counts show attention, not usability; the more useful evidence for capability judgment is the official mechanism description plus a small number of explanatory community posts.

Its practical role is not an “automatic prompt-writing tool,” despite one social post framing it that way. A more accurate analogy is a benchmark-informed routing layer for model and provider selection inside OpenRouter. The output it helps improve is not copy or prompts, but steadier function calling, agent step execution, and less manual model/provider switching. The available evidence also suggests a narrow focus on tool-calling reliability rather than being a general chatbot booster or a standalone agent framework.