Fianza

Getting early access is straightforward. Just dropping an email gets you on the testnet. This convenience lowers the barrier to entry significantly for developers wanting to build.

I am looking at the website to understand how this actually works in practice. The text mentions that agents can draw against a credit line and repay as they earn, which sounds promising. However, it is not entirely clear what happens if the agent defaults. The text says the machinery bites back and there is socialized loss, but it does not explain how that loss is distributed among lenders or what the specific penalties are. Without more detail on the risk model, it is hard to assess the safety of lending capital. Also, the exact fees for borrowers are not spelled out. It seems like a interesting experiment, but I would want more concrete information before committing.

The range of tools is impressive. They offer dashboards for lenders and borrowers, plus SDKs for npm and pip, making it easy to integrate with AI agents on Stellar.

Fianza positions itself as a provider of on-chain credit specifically designed for AI agents, operating on the Stellar network with a focus on uncollateralized USDC lending. The core mechanism involves a continuous loop where an agent earns revenue, which is then used to prove its financial health through on-chain data. This proof feeds into an adversarial scoring engine that discounts unreliable income, such as sybil activity or circular funding, to determine an effective revenue figure. Based on this scored revenue, the system extends an isolated credit line, allowing the agent to draw funds when needed, such as to pay for a paywall, and repay as new revenue lands. The process is designed to be automated, with underwriting running in seconds rather than traditional banking cycles. For lenders, capital is supplied into these isolated vaults to earn yield from repayments. The platform highlights a specific agent example where revenue from x402-priced calls is indexed, and the discount engine identifies distinct payers while flagging circular funding. While the technology seems innovative, relying on zkTLS proofs and on-chain observability, it is currently in testnet phase. The team is building an SDK for agents and offers dashboards for both borrowers and lenders, suggesting a growing ecosystem. However, the exact terms of the credit lines and the precise mechanics of the 'socialized loss' upon default remain to be fully explored as the platform matures beyond its early access stage.

It is refreshing that they offer a testnet early access program where you simply drop an email to get onboarded. This removes friction and makes it easy to start exploring their on-chain credit system immediately.