Helicone

Helicone combines an AI gateway with request tracing, cost tracking, caching, rate limits, and evaluation tools. It works as a proxy in front of model providers, supports self-hosting, and exposes operational data to agents through an MCP server for debugging and optimization workflows.