Cadreen – 记忆、治理、自我修复与执行为一体的系统
嗨,HN——我正在发布Cadreen的研究预览。
Cadreen是智能工作流的基础设施——一种能够记忆、决策、行动、请求权限并留下审计记录的软件。它包括API、SDK、CLI工具、内存、治理、工具执行、审计记录和聊天访问。
这个想法是,有用的AI系统不仅仅需要模型调用。它们需要协调:记住什么、何时行动、何时请求权限、使用哪些工具,以及如何留下审计记录。
现在可用:
- API和文档
- TypeScript、Python和Go SDK
- CLI
- Cadreen可作为聊天界面中的模型/提供者使用(即将推出的BYOK将实现管理模型选择)
- 接下来推出:“完成任何事情”,一个供非开发者直接使用相同基础设施的聊天界面。
欢迎反馈——特别是来自构建代理、协同助手或AI工作流系统的人的反馈。
https://github.com/timothy-billingrails/cadreen-sdks
查看原文
Hi HN — I’m launching the research preview for Cadreen.<p>Cadreen is infrastructure for intelligent workflows — software that can remember, decide, act, ask permission, and leave an audit trail. It includes APIs, SDKs, CLI tooling, memory, governance, tool execution, audit trails, and chat access.<p>The idea is that useful AI systems need more than a model call. They need orchestration: what to remember, when to act, when to ask permission, what tools to use, and how to leave an audit trail.<p>Available now:<p>- APIs & docs<p>- TypeScript, Python, and Go SDKs<p>- CLI<p>- Cadreen usable as a model/provider in chat interfaces (managed model selection with BYOK coming soon)<p>- Coming next: “Accomplish Anything,” a chat interface for non-developers to use the same infrastructure directly.<p>Feedback welcome — especially from people building agents, copilots, or AI workflow systems.<p>https://github.com/timothy-billingrails/cadreen-sdks/