Every team shipping AI features gets the same invoice: one number per provider. It says nothing about which feature spent it, which model, or which customer it was for. We built Alye to answer those questions, and to tell you the minute a call starts failing.
The problem we kept running into
Once a product has more than a couple of AI features, the provider dashboard stops being useful. Several services share one API key. Retries hide timeouts. A rate limit at 2 a.m. degrades answers for an hour and nobody notices. And when finance asks what the new assistant costs per customer, the honest answer is a spreadsheet and a guess.
Prices make it harder. The same model costs more in an Azure data zone or a Bedrock regional profile, and prompt caching changes the bill again.
What Alye does
- Cost by service, model and customer. Every call is priced from a versioned price list, including regional and marketplace prices, and your own prices for fine-tunes.
- Traces. Each request becomes a trace with its model calls, retries and errors, so you can see what actually happened.
- Alerts. Spend limits, monthly forecasts, error rates, slow calls and services that stop reporting, checked every minute and sent to email, Slack or Microsoft Teams.
- Customer budgets. Tag requests with your customer ID to see what each account costs and set a monthly budget for it.
How it works
Alye is an SDK, not a proxy. It wraps the official OpenAI, Anthropic and Google clients inside your service and sends records in the background, so your requests go straight to the provider and never wait on us.
pip install alye import alye alye.init(service="support-reply") # set ALYE_API_KEY in the environment
Node.js works the same way with npm install alye, and there are integrations for Express and NestJS. Services that already emit OpenTelemetry GenAI spans can send them to Alye directly.
Prompts and responses are not sent unless you turn that on. When you do, personal data is masked inside your service before anything leaves it.
What Alye deliberately doesn't do
Alye doesn't run evaluations, manage prompts, cache responses or route traffic. Good tools exist for those, and many teams will run Alye next to them. We focus on one job: knowing what production AI costs and when it breaks.
Try it
The Free plan records 500,000 calls a month. Team is $49 a month for 10 million calls with unlimited members, and we never charge per seat. See pricing, read the docs, or start with one service. Your first trace should appear in about five minutes.
Questions or feedback: info@alye.ai.