How Revrag tracks costs across 80+ AI agents

The challenge

Revrag builds and deploys conversational AI agents for enterprise clients, handling customer calls across health, motor, wealth, and POS verticals. They operate 80+ agents concurrently, and each one generates its own consumption trail measured in call minutes, speech-to-text usage, and LLM inference.

Every one of those signals maps to a different cost driver and a different end customer. At that scale, the reporting layer became the bottleneck rather than the infrastructure itself.

• No visibility into per-agent cost
• Manual invoicing via spreadsheets
• AWS costs couldn't be split by agent or customer
• STT usage discrepancies went undetected
• Billing lagged real-time consumption by weeks

The solution

Revrag connected each agent in its pipeline to Paygent, reporting the specific usage signal that drives that agent’s cost. Paygent meters each event, prices it against the customer’s plan, and attributes it to the right account.

Because every agent type carries a different cost profile, they are tracked and billed independently rather than rolled into a single platform line item.

The results

Cost tracking now runs in real time against live consumption, and each customer receives separate cost attribution and its own invoice. Billing activates the moment an agent goes live, with no manual handoff between the engineering and finance sides.

• Real-time cost tracking per agent
• Automated monthly invoice generation
• Per-customer pricing and volume tiers
• Agent enable/disable triggers billing automatically
• Minimum commitment enforcement at a 10,000-minute floor
• GST, TDS, and credit notes handled on generated invoices