可随时重新打开的智能体运行记录。

Switch on session capture and every run through the gateway becomes a trace you can open, export, share and re-ask step by step on another model — and the open-source CLI records the shell, filesystem and MCP calls a gateway never sees.

Once capture is on, every run carries the routing and cost truth: every turn, every model call, and what each one actually cost.
步骤级假设推演：将某一轮对话重新发送至其他模型，并将各模型的回答、延迟及费用并排对比。
将任意运行导出为 orca-trace v0 归档文件，或通过可设置有效期、可随时撤销的分享链接转交他人。

两个半段，一种追踪格式。

本地 orca record 命令可捕获网关所无法触及的内容：Shell 环境、文件系统及 MCP 调用。网关则捕获本地端永远无法获取的信息——每一个密钥、每位团队成员、每个 CI 任务——并附带路由决策和实际费用。两者均写入相同的 orca-trace 归档格式，因此无论从哪一侧生成的运行记录，打开方式完全一致。

提出那个事后才想到的问题。

Find the turn where it went wrong and re-ask that one step on a different model. The comparison runs through your own gateway on your own keys, so what comes back is what you would actually have shipped — with the token bill next to it.