ZTAgent.ai develops open-source security infrastructure for AI agents. Its Apache-2.0 Python gateway combines OIDC authentication, registered-tool controls, OPA policy integration and HMAC-chained audit records. Developers can apply configured security policies to model and tool calls routed through the gateway. The project includes public code, tests, documentation and examples, with Victor Fang as project lead. A Celo extension is planned to add native/ERC-20 transaction intent controls, Self Agent ID verification and transaction-linked audit evidence. These Celo features are proposed work; no existing Celo deployment or usage is claimed.
Make agent security inspectable and reusable through open-source identity checks, policy enforcement and audit tools.
AI agents can invoke tools and access services with significant permissions. Developers need a consistent way to verify identity, restrict actions and trace decisions across agent workflows.
ZTAgent provides an Apache-2.0 Python security gateway with OIDC authentication, registered-tool controls, OPA policies and HMAC-chained audit records. A planned Celo extension will add transaction intent checks, Self Agent ID verification and transaction-linked audit evidence.