SENTRY | Docs

Introduction

SENTRY starts from a simple premise: agent authority should be defined before execution, not inferred after an action occurs. The interface expresses permissions, boundaries and policy decisions through one readable surface.

Core sequence

request → policy → preflight → execution → record

Core concepts

Agent

An automated actor that proposes an onchain request through an operational workflow.

Policy

A set of explicit rules that determines whether the request fits the agent’s allowed scope.

Execution context

The network, target, asset, amount and action information attached to the request.

Record

A structured outcome that captures the decision context and resulting status.

Policy example

A treasury policy can be represented with a transaction ceiling, daily operating budget and approved target list.

policy treasury_01 { network = ARC assets = [USDC, WETH] max_transaction = 250 USDC daily_budget = 1000 USDC target = ALLOWLIST_ONLY preflight = REQUIRED }

Operating model

SENTRY is designed to give operators a clear policy view before execution and a clear record after execution. The simulator mirrors that logic locally in the browser so the decision model can be inspected without connecting a wallet.

Reference the SENTRY operating model.