Feyn公司推出了Critic,一款AI驱动的代码审查平台,使开发人员能够直接与生成代码的AI Agent进行交互1。该平台通过Codex或Claude Code插件运作,并通过MCP提供全部功能1。用户可以对Agent生成的代码提出问题并获得解释,而AI Agent则能够编写叙述性描述来解释变更的背景、假设和决策逻辑,特别是对复杂代码段的注解1。
Critic内置了多项安全机制,包括分叉会话功能,该功能可移除写入工具,使会话仅能回答关于代码变更的问题1。在Feyn公司的生产前测试中,该工具已帮助发现了多个问题,包括模型未重用现有工件和误诊问题等情况1。该平台目前免费提供,用户可访问https://critic.run获取使用[1](#source-1)。
Feyn has launched Critic, a code review platform that enables developers to directly interact with the AI agent responsible for writing their code.1 The tool operates through Codex or Claude Code plugins via MCP, providing a novel approach to reviewing code generated by artificial intelligence.1
Critic allows AI agents to annotate code and provide contextual information about changes, including background narratives, underlying assumptions, and decision rationale for complex implementations.1 Users can pose questions about agent-generated code and receive explanations directly from the agent that created it.1 The platform includes a security feature whereby forked sessions remove write tools, restricting interactions to questions specifically about the code changes.1
According to Feyn, Critic has already helped identify several issues before production deployment, including instances where the model failed to reuse existing artifacts and cases of misdiagnosis.1 The platform is currently available free of charge at https://critic.run.[1](#source-1)
评论
还没有评论,欢迎留下第一条。