What is agentkredit?

AgentKredit is a blockchain based system that gives autonomous AI agents, such as trading bots or automated treasury managers, a way to spend money on their own within strict, provable limits. The project targets a problem the author describes directly: today an autonomous agent either holds a private key with unlimited access, which is risky if it misbehaves or gets compromised, or it needs a human to approve every transaction, which defeats the point of automation. AgentKredit tries to sit between those extremes. Agents build up a reputation by having their actions attested to on chain, and that reputation converts into a credit line with a spending limit tied to their reputation tier. To prove things about their own behavior, such as maintaining a certain success rate, without revealing private details of what they actually did, agents use zero knowledge proofs built with the Circom language and the Groth16 proving system. Spending is then enforced through on-chain rules including daily caps, whitelisted recipients, and time-limited session keys, so an agent can act independently while staying inside a policy a human or protocol has set. The system is built as a set of smart contracts on a blockchain called BOT Chain, developed and tested using the Hardhat framework and Solidity. Supporting off-chain services include an attestation oracle that watches on-chain events and signs attestations, and a prover service that generates the zero knowledge proofs locally from an agent's private action log. According to the README's own status tables, the core trust registry, the zero knowledge proof layer, and the credit and policy enforcement contracts are built and tested, while a dashboard, a software development kit for agents, an event indexer, and an actual testnet deployment are still marked as not yet done.