A billing and margin layer for self-hosted LLMs. It meters every token, bills it through PayPal, and an agent keeps each customer and model profitable. It is a billing and margin layer for teams that self-host LLMs. An OpenAI-compatible gateway sits in front of your models and meters tokens, latency and GPU-seconds for every customer. Usage flows into PayPal’s Usage-Based Billing API, which turns it into invoices and collects payment automatically. An AI agent then joins that revenue with real GPU costs to show margin per customer and per model. It flags unprofitable customers, idle GPUs and SLA breaches, proposes fixes such as repricing or switching to a quantized model, and applies them after approval in Slack. Built with PayPal, Kernel, Render Workflows and Astropods.
Built With
- gpu
- kernel
- llm
- paypal
- render
- sglang
