Two weeks ago, Blockster covered agents applying for their own credit lines. Now t54 is reporting another milestone in what happens when software starts spending.

On October 1, t54 Labs announced that it had surpassed 10 million agent-initiated transactions on the XRP Ledger. The company also reported more than 169,000 applications for Claw Credit, with roughly $650,000 in credit extended.

When Blockster covered Claw Credit on September 17, t54 reported more than 135,000 applications and over $500,000 in credit extended. The latest figures show the company continuing to expand its experiment in giving software controlled spending power.

For businesses selling data, computing power and digital services, that creates a new kind of customer: an agent with a task to finish and a budget to work with.

What Are Agents Actually Buying?

An agent researching companies might need access to a paid financial database. A coding assistant could buy computing time. Another could pay a specialist service to check information before completing a report.

These purchases can happen through x402, a payment standard that lets software receive a price, pay for access and continue working. XRP Ledger’s implementation, contributed by t54, supports payments using XRP or Ripple’s dollar stablecoin, RLUSD.

Instead of sending a person through a checkout every time it needs something, the agent can purchase a resource within its permitted spending limits.

“Ten million transactions on XRPL demonstrates the speed with which agents are becoming genuine economic actors, well beyond software that simply recommends what a human should do next,” said Chandler Fang, co-founder of t54.

The transaction count measures activity. The announcement does not disclose the total dollar value of those 10 million transactions or how many individual agents generated them.

From Credit Applications to Spending History

Claw Credit tackles the funding side of that activity.

As we explored in our earlier coverage, the model resembles a corporate spending account for software. Approved agents receive controlled credit lines for services they need, while their behavior and repayment history help determine future access to credit.

The product supports XRPL, Solana and Base. Its risk engine considers who operates an agent, how it spends and whether it repays on time.

For a business running several assistants, the appeal is practical. A research agent could have a budget for data, while a developer agent receives an allowance for computing services. Neither needs unrestricted access to the company’s wallet.

The latest application and credit figures suggest growing interest in that model. Applications, however, are separate from approvals, and credit extended is separate from money spent or repaid. Those measures will help show how the business develops.

Trust Before Money Moves

As agents gain spending power, companies are building controls around what they can do. This week, Blockster covered NVIDIA’s work with more than 100 organizations on agent-safety technologies, including restrictions on access to tools, accounts and systems.

t54 tackles the financial side: verifying who an agent represents, setting spending limits and checking risk before money moves.

With Claw Credit, spending behavior and repayment history also help determine how much credit an agent can access.

For a business, the appeal is practical. A research assistant could buy data and a coding agent could pay for computing time, each working within an agreed allowance.

Small Payments, Large Transaction Counts

A September AWS case study, co-authored by t54 team members, reported more than 20 million agent-initiated transactions through its trust layer on Amazon Bedrock AgentCore payments.

Those payments ranged from $0.001 to $0.01—one-tenth of a cent to one cent. An agent can make many such purchases while completing a task, helping explain how transaction counts climb quickly. The system checks payments before they proceed, and the agent stops when its spending allowance runs out.

For businesses selling data or computing services, this opens up a different way to charge: individual requests and tasks purchased automatically as needed. The next test is whether those small payments turn into repeat customers and steady revenue.

Reporting by Lidia Yadlos

AI AgentsClaw Creditt54 Labsx402 paymentsXRPLXRP Ledger