Respan

Respan is an AI gateway that routes requests to more than 1,000 models through one API and adds observability and evaluation on top. Developers building agents use it to trace runs, set fallbacks, cache responses and score outputs.

You point your existing SDK at Respan's endpoint and switch models by changing one value, with automatic fallbacks when a provider errors or rate-limits. Response caching cuts cost on repeat calls, and spend limits keep usage in check.

Traces, monitors and red-team tests cover every agent run, and LLM-as-judge evaluators score outputs against datasets. The gateway has been released as open source, and paid plans add unlimited datasets and evaluators, SOC 2 reports and an optional HIPAA add-on.