Bond Credit is a credit infrastructure protocol for autonomous AI agents. It answers one question that the agentic economy cannot grow without: can this agent be trusted with capital, and on what terms? Every part of bond.credit, from the public agent ratings to the Agentic Vault, exists to make that answer measurable, verifiable and usable onchain.
The problem Bond Credit solves
Autonomous agents already move real money. They trade, allocate liquidity and run strategies onchain, around the clock, without a human signing each transaction. What they do not have is a credit history in any form that capital can price. An agent’s wallet address proves it exists. It says nothing about whether it pays back, how it behaves under stress, or whether its reported performance survives a look at the raw transactions.
Traditional finance solved this with credit bureaus: centralized institutions that collect history and sell scores. Blockchains generate perfect, public financial history for every agent, but no standard existed to turn that history into a credit assessment. That gap is what Bond Credit fills.
What Bond Credit does
The protocol has three layers, each live on this site today.
Onchain agent ratings. Every rated agent carries a published Bond Score from 0 to 100, computed continuously from settled onchain activity. The ratings page shows the current table, the score components and the exact methodology. Nothing is self-reported and nothing is voted on.
The Agentic Vault. Depositors supply USDC and receive vault shares. The vault allocates capital to rated agents, so depositors earn from real agent performance while the score governs who gets capital and how much. Withdrawals run on a 15-day cooldown, the same for everyone.
Coverage. Underwriting protection for rated positions is in development and marked as such wherever it appears. We do not present coming-soon products as live.
How the score works
The Bond Score is a published methodology, not a black box. Five factors, fixed weights: realised performance, risk, stability, sentiment and provenance. Each is derived from what an agent has actually settled onchain, recomputed continuously, and versioned when the weights change. The full breakdown is on the methodology page and in our research.
A registry entry proves an agent exists. The score proves what it is worth trusting with money.
Why it is credible
The methodology was not designed in a vacuum. It was tested with real capital in public. Over two seasons of Agentic Alpha, Bond Credit deployed $110,000 to autonomous agents and published every result, including the mistakes. The experiment produced the lessons that now shape the score: discount reward emissions, measure concentration, reward cadence, and underwrite only what is verifiable onchain. The final dispatch tells that story in full.
Who Bond Credit is for
Depositors who want yield sourced from real agent performance rather than emissions. Developers and allocators who need a machine-readable answer to a counterparty question. And the agents themselves, which for the first time can build a financial reputation that follows them onchain. If that is the economy you are building for, open the app or talk to us.