Skip to main content
TrustLine is a revenue-underwritten, uncollateralized USDC credit protocol for AI agents on Stellar. An agent proves what it actually earns — on-chain x402 revenue, optionally backed by a zkTLS-verified off-chain figure — and gets a credit line sized and priced against that proof. It draws and repays autonomously. No collateral, no human in the loop, no wallet-age heuristics.
Status: live, working prototype on Stellar testnet. The core loop (earn → prove → underwrite → borrow → repay → lender yield) runs end-to-end on real testnet infrastructure. Not yet on mainnet — see Roadmap for what that requires.

​
The five-second mental model

1

Earn

An agent earns real USDC revenue (x402 payments, optionally + a zkTLS-proven off-chain figure).
2

Prove independence

TrustLine’s underwriting engine checks that revenue is genuinely independent (not the agent quietly paying itself), scores it, and publishes the score on-chain.
3

Price

An on-chain credit_line contract derives a credit limit and APR from that score.
4

Fund

Lenders deposit USDC into the agent’s isolated vault — exposed only to that one agent’s default risk, never pooled across agents.
5

Borrow & repay

The agent borrows against its line and repays as it earns. On-time repayment ramps the limit up over time; a default collapses it and socializes the loss to that vault’s lenders via a first-loss reserve.

​
What do you want to do?

What & why

Understand why this exists and the core thesis.

Architecture

See how the pieces fit together — contracts, backend, SDK.

How the credit engine works

Understand exactly how revenue is judged and scored.

Onboarding kit

Build an AI agent that earns credit on TrustLine.

SDK reference

The full agent-facing API — JavaScript/TypeScript and Python.

Contract addresses

Look up a deployed contract ID.

Live app

0xtrustline.online

Live underwriter

Paste any address, watch a live verdict.

GitHub

TechnicallyKiller/TrustLine

Roadmap

See what’s shipped vs. planned.
The backend API is at https://trustline-rpxt.onrender.com.

​
Honest status, up front

This is a genuine, adversarially-tested testnet prototype — not a production lending product. The credit-risk engine, anti-Sybil independence model, and safety rails are real and tested (fuzzed contracts, a live on-chain default run, a synthetic + live attacker catalog). What’s explicitly not yet true: nothing has touched real money or a real adversary, the score-signing key is a single trusted signer (not yet decentralized), and external adoption is still near zero. See Roadmap for the plan to close each of these.