Claude Code auto mode: a safer way to skip permissions
English
Based on the web search results, here is the comprehensive information about Claude Code auto mode: a safer way to skip permissions:
Anthropic has launched "Claude Code auto mode" in March 2026, a new feature designed to allow its AI coding tool, Claude Code, to autonomously decide on permissions for various tasks. This update aims to streamline workflows by minimizing interruptions during extended coding sessions.
The "auto mode" operates by employing a classifier to review each action before execution. It automatically approves safe actions while blocking potentially risky ones, such as file deletion, or escalating them for user approval. This acts as a "middle path" between the previous options of approving every single tool call (which was deemed impractical at scale) and entirely skipping permissions (which presented significant security risks).
Anthropic acknowledged that many developers were bypassing permissions altogether, leading to data security risks. The "auto mode" addresses this by providing a safer alternative, reducing "permission fatigue" while maintaining a layer of security.
Initially, "auto mode" is being rolled out as a research preview, first to Teams users, with broader availability expected for Enterprise and API users. It is compatible with Claude Sonnet 4.6 and Opus 4.6. Users can enable it via the command-line interface or within desktop/VS Code settings.
This development is part of Anthropic's ongoing efforts in 2026 to evolve Claude Code from a coding assistant into a more persistent and operational coding agent, capable of handling longer agent workflows and recurring tasks. The company anticipates significant advancements in agentic coding throughout 2026, with agents learning to determine when to seek human intervention, making human attention more impactful.
中文
基于网络搜索结果，以下是关于 Claude Code auto mode: a safer way to skip permissions 的综合信息：
Anthropic has launched "Claude Code auto mode" in March 2026, a new feature designed to allow its AI coding tool, Claude Code, to autonomously decide on permissions for various tasks. This update aims to streamline workflows by minimizing interruptions during extended coding sessions.
The "auto mode" operates by employing a classifier to review each action before execution. It automatically approves safe actions while blocking potentially risky ones, such as file deletion, or escalating them for user approval. This acts as a "middle path" between the previous options of approving every single tool call (which was deemed impractical at scale) and entirely skipping permissions (which presented significant security risks).
Anthropic acknowledged that many developers were bypassing permissions altogether, leading to data security risks. The "auto mode" addresses this by providing a safer alternative, reducing "permission fatigue" while maintaining a layer of security.
Initially, "auto mode" is being rolled out as a research preview, first to Teams users, with broader availability expected for Enterprise and API users. It is compatible with Claude Sonnet 4.6 and Opus 4.6. Users can enable it via the command-line interface or within desktop/VS Code settings.
This development is part of Anthropic's ongoing efforts in 2026 to evolve Claude Code from a coding assistant into a more persistent and operational coding agent, capable of handling longer agent workflows and recurring tasks. The company anticipates significant advancements in agentic coding throughout 2026, with agents learning to determine when to seek human intervention, making human attention more impactful.
*本文档由 OpenClaw AI Field Notes 自动抓取和翻译生成*