PurrCat 文档 PurrCat documentation

先把 PurrCat 装起来Install PurrCat first

一条命令装好，再启动桌面端；接口默认监听 http://127.0.0.1:8000。 Install with one command, then launch the desktop app. The API listens on http://127.0.0.1:8000 by default.

什么是 PurrCat？What is PurrCat?

一个基于信息熵不可逆性理念设计的 Agent 底座。 An agent foundation designed around the irreversibility of information entropy.

信息一旦丢失，就无法保证被完整还原。就像一张被打上马赛克的图片，即使最强大的模型，也只能根据剩余信息推测原貌，而无法保证还原出真实的原图。 Once information is lost, nothing guarantees it can be fully restored. Like a picture with a mosaic over it: even the strongest model can only guess at what was there from what remains — it cannot promise the original back.

人与 Agent 的沟通也是如此。你表达的需求、任务中的背景、执行过程中的发现，都可能在传递中遗漏、压缩或失真。模型能力越强，并不意味着它就能自动知道你真正想要什么。 The same goes for how you talk to an agent: the needs you express, the context behind a task, the discoveries made while running it — any of them can be dropped, compressed, or distorted on the way. A stronger model does not automatically know what you actually want.

我们尊重信息熵不可逆这一事实。因此，PurrCat 不仅致力于提升 Agent 的任务交付能力，更致力于打造你与 Agent 之间更方便、更高效的信息信道。 We take the irreversibility of information entropy as a given. So PurrCat works not only to raise the agent's task delivery, but to build a more convenient and more efficient information channel between you and your agent.

让你的意图更容易表达，让必要的信息更准确地流动，让 Agent 更贴近你真正想要的结果。 Make your intent easier to express, let the necessary information flow more accurately, and bring the agent closer to the result you actually want.

Build an Agent That Works Your Way. Build an Agent That Works Your Way.

为什么选择 PurrCat？Why PurrCat?

我们相信，好的 Agent 不应该只追求更强的模型，还应该拥有更好的信息流转方式。 We believe a good agent should not only chase a stronger model, but also carry a better way for information to flow.

模型会越来越强，工具会越来越多，Agent 能承担的任务也会越来越复杂。但有一件事不会因此自动消失：Agent 无法凭空获得你没有表达、系统也无法可靠推断的信息。 Models will keep getting stronger, tools will keep multiplying, and the tasks an agent can take on will keep getting more complex — but one thing will not disappear on its own: an agent cannot conjure information you never expressed and the system cannot reliably infer.

因此，我们希望 Agent 的进步，不只体现在模型能力上，也体现在人与 Agent 之间的信息如何表达、组织、传递和留存。 So we want the agent's progress to show up not only in model capability, but in how information between you and the agent is expressed, organised, passed on, and kept.

这就是 PurrCat 的设计出发点：提升任务交付能力，打造更好的信息信道，让 Agent 不只更强，也更懂得按照你的方式工作。 That is where PurrCat's design starts: raise task delivery and build a better information channel, so the agent is not only stronger — it also knows how to work your way.