Respan

Respan (formerly Keywords AI) is a unified LLM engineering platform that routes traffic through a single gateway reaching 1,000+ models, then captures every call as a trace while tracking spend, errors, and latency and scoring outputs with evals. It's built for AI/LLM engineers and product teams running LLM apps and agents in production who need one layer for routing, observability, and evaluation instead of stitching separate tools together. Its main differentiator is combining an OpenAI-compatible gateway (with fallbacks, caching, and budgets) with deep span-level tracing and composable evaluators in the same product.