Tell the agent how to spend money in plain English — it parses that into structured rules, then explains every decision · 用自然语言告诉 Agent「怎么花钱」，它会解析成结构化规则，并在每次决策时给出洞察与解释
Two places the AI takes part · 两处 AI 参与：① Type one sentence below to set your constraints — the AI parses it into structured rules. · ① 在下面输入一句话设定约束，AI 会把它解析成结构化规则。 ② In the Insight / Explanation panel, try a purchase — the AI first works out your budget position, then gives an insight, then an actionable explanation for the verdict. · ② 在「消费洞察 / 决策解释」区试一笔采购 —— AI 先算预算位置，再给洞察，最后给可执行的决策解释。
① Set constraints in plain English · 用自然语言设定约束
e.g. "SaaS budget 200 dollars a month, anything over 50 ask me first, only Notion and Figma" · 例：「每月 SaaS 预算 200 美元，单笔超过 50 要先问我，只允许买 Notion 和 Figma」
② Insight / Explanation — try a purchase · 消费洞察 / 决策解释（试一笔采购）
e.g. "Buy Notion Pro for 10 dollars" or "Buy Slack Pro for 15 dollars" (blocked by the allowlist, with an explanation) · 例：「帮我买 Notion Pro，10 美元」或「买 Slack Pro，15 美元」（会被白名单拦截并给出解释）
💡 Insight 消费洞察
🧭 Explanation 决策解释
③ Tool trace & price comparison · 工具轨迹与比价结果
Every tool call the model made, in order — the visible proof that the agent reasoned instead of just classifying. · 模型按顺序发起的每一次工具调用 —— 这是「AI 在推理而非只做分类」的可见证据。
No tool calls in this decision — the rule engine decided it directly. · 本次决策没有工具调用（规则引擎直接判定）。