Mandato: Protocol-Level Enforcement of Digitally Signed Mandates on AI Agent Actions

A new paper on arXiv (2608.14074v1) introduces Mandato, a governance proxy designed to enforce digitally signed mandates on AI agent actions at the protocol level. The system addresses the lack of infrastructure constraining AI agents to authorized actions when they interact with external systems via standardized tool-calling protocols like the Model Context Protocol (MCP). Currently, authorization logic resides in application code, is unsigned, and not independently auditable, resulting in logs lacking evidentiary value. Mandato evaluates every tool call against a mandate chain, blocking non-conforming calls in real time and recording each decision in an append-only, hash-chained audit log. The paper details the architecture and implementation of this protocol-level enforcement mechanism, highlighting its potential to provide verifiable authorization and auditability for AI agent actions.