Zero Trust for the LLM
and the upstream. 对大模型、上游，
零信任。

Untrusted in. Verified out. 不可信进，验证后出。

Sieve assumes every external input reaching your agent is hostile until proven otherwise, and gates every irreversible thing your agent can do. Between the two sits one local checkpoint — policy plus your intent — that nothing crosses unverified. Sieve 默认抵达 Agent 的每一个外部输入都是敌意的，直到被证明无害；并为 Agent 能做的每一件不可逆的事设置闸门。两者之间，只有一个本地检查点——策略加你的意图——任何东西未经验证都无法通过。

Trust nothing on the wire — not the LLM, not the upstream. 链路上谁都不信——不信大模型，也不信上游。

Coding agents now sign transactions, move funds, and deploy contracts on your behalf. The prompt leaves your machine with your secrets inside it; the LLM sends back tool calls that execute without a second look. Zero-trust means assuming either end can be wrong — and putting one checkpoint on the only wire you actually control. 编码 agent 如今会代替你签名交易、转移资金、部署合约。提示词带着你的密钥离开本机；大模型把工具调用发回来，未经第二眼便执行。零信任，意味着假设两端都可能出错——并在你唯一真正掌控的那条链路上，设一个检查点。

Built for crypto's irreversible actions. 为加密世界中不可逆的操作而生。

The gateway is general; the ruleset is not. Crypto is where Sieve goes deepest — the differentiator no general-purpose guard matches — with detection tuned for the actions you can never take back. 网关是通用的，规则集却不是。Crypto 是 Sieve 下探最深的地方——通用防护无法企及的差异化护城河——检测专为那些永远无法收回的操作而调校。

Up in one command. 一条命令即可启动。