# Understand the usage consuming your free-credit budget.

Every free trial spends real compute. See which signups and runs come from repeated automated accounts, and connect that evidence to the limits, payment checks and reviews you already run.

For AI and SaaS teams offering free credits or costly operations.

Status: Discuss an integration

## Free credits are paid for in compute.

Inference, extraction and search cost money per call. A script that opens trial after trial spends your budget without ever becoming a customer.

Automation alone is not the problem: your paying customers’ agents call your API all day. The question is which usage is eligible for free credits, and that is your policy.

## Where $50,000 of free credits went

- Signups: 10,000
- Free credits granted: $50,000
- Repeat-trial pattern: 1,500
- Credit cost at stake: $7,500

Illustrative example. 10,000 signups with $5 of credits each; 15% with a repeat-trial pattern. Hypothetical.

## The trial funnel, two ways.

| | Reported | With evidence |
|---|---|---|
| Trials | 10,000 | 8,500 |
| Credit cost per real trial | $5.00 | $5.88 |
| Budget on repeat trials | $0 | $7,500 |

Illustrative example.

## Your numbers

At stake = signups × share repeat-trial × credit. Cost per real trial = credits granted ÷ trials not automated.

## Three accounts, three different answers.

- Paying customer’s agent: Authorized: keep serving
- New trial, person at a browser: Legitimate: grant credits
- Fifth trial from one script: Suspected abuse: review
- Your control: Payment check before credits

Same automation, different decisions. Your policy draws the line.

## How it fits

1. **Observe the signup.** Label web signups and credit claims in the browser.
2. **Authenticate API use.** Use account and service-event evidence for API calls; a browser script does not see them.
3. **Join metered usage.** Connect accounts to the resource units they consume.
4. **Apply your controls.** Review, require payment, budget or rate-limit, as you decide.

## What your team can do

- **Hold credits for review.** Delay free credits for suspected repeat trials.
- **Ask for a payment method.** Require a card before costly operations.
- **Measure marginal cost.** Track real resource cost, with estimates labelled.

## Before you start

### Isn’t all API traffic automated?

Yes. Human versus bot is not the eligibility rule; repeated trial abuse is.

### Does the browser script see my API calls?

No. Server calls need authenticated account and service evidence.

### Is this a gateway or entitlement system?

No. It informs the controls you run; we do not promise a released gateway or cross-account identity detector.

### How do I compare it with my rate limits?

Measure what it adds beyond the rate limiting already in place, after false positives.

[Discuss your usage-abuse problem](mailto:support@doubleagent.so?subject=Use%20case%3A%20Free-credit%20and%20API%20abuse)

- [Double Agent Identity](https://doubleagent.so/identity/)
- [Double Agent for SaaS](https://doubleagent.so/for/saas/)
- [All use cases](https://doubleagent.so/use-cases/)