LiteLLM
self-hosted proxyA proxy you run yourself. Its docs say it tracks spend "for keys, users, and teams across 100+ LLMs". Tag-level spend is an enterprise feature.
Best for: teams that want the gateway on their own servers. Connect it with LiteLLM cost tracking in LLMtrack.
OpenRouter
hosted routerOne API and one bill across many providers, with automatic fallback when a provider errors. It charges a 5.5% fee ($0.80 minimum) when you buy credits and says there's no markup on inference prices.
Best for: small teams that want many models without many accounts. Connect it with OpenRouter cost tracking in LLMtrack.
Cloudflare AI Gateway
hosted gatewayCaching, rate limiting, retries, model fallback, logs and analytics including cost. Cloudflare lists it as available on all plans.
Best for: apps already on Cloudflare. Connect it with Cloudflare AI Gateway cost tracking in LLMtrack.
LLMtrack
cost layer, not a gatewaySits beside any of the three above, or works alone through the Node and Python SDKs. Prices every request component by component from the AI API pricing sheet for 422 models, then shows cost per feature, margin per customer and the day you'll pass your budget.
Free plan: 3,000 requests a month, every feature. Pro: $11.99 a month.