Respan

What it does

Respan is an AI gateway that routes requests to many models through one API. It adds tracing, monitoring, evaluations and security testing for AI agents. You can set fallbacks, cache responses and apply spend limits. The platform also scores outputs, tracks cost and latency, and masks PII. A free account, demo booking and documentation are available.

The gateway reaches models across major providers and lets you switch by changing one word. Fallbacks move to the next model when one errors or rate-limits. Response caching serves repeat calls, and budgets and rate limits can be set per key, per customer or org-wide. Tracing records each LLM call, tool run, retrieval and agent turn as a span with input, output, latency and cost. Evaluations combine LLM judges, code checks and human review, and can run on sampled production traffic. Monitoring shows requests, errors, cost, latency and tokens with alerts to Slack, email or a webhook. Security features include PII masking and red teaming. SDKs are listed for Python, JS/TS, OpenAI, OpenAI Agents, Vercel AI, Mastra and LangChain.