SINT Protocol
An open protocol and TypeScript reference stack that authorizes and logs AI agent actions with real-world consequences, using Ed25519 capability tokens, tiered human-approval rules and a tamper-evident, hash-chained evidence log.
SINT Protocol sits between an AI agent's intent and its real-world execution. Before a governed tool call, robot command, industrial write or actuator movement runs, the request passes through a PolicyGateway that checks a capability token for authorization, applies a T0 to T3 approval tier that escalates physical or irreversible actions to a human operator, checks whether the token has been revoked, and logs a tamper-evident record of the decision.
Ed25519 capability tokens bind an issuer, subject, resource, action, expiry and delegation chain, and travel with constraints such as velocity, force, geofence, budget and rate limits instead of those limits living in separate application configuration. Every decision can be written to a SHA-256 hash-chained evidence ledger with portable proof receipts. The protocol is explicitly an enforcement layer in front of existing systems - MCP, A2A, ROS 2, MAVLink, MQTT and OPC UA - rather than a replacement for them, and a separate mission-authority extension covers autonomous air, ground and maritime platforms without selecting targets or generating effects.
It ships as an open protocol specification plus a TypeScript reference implementation.