On March 25, 2026, Anthropic announced the launch of the 'Auto Mode' permission management feature in Claude Code, offering developers an intermediate solution between the default conservative configuration and completely skipping permissions. This mode allows tasks to run for extended periods with reduced manual intervention, featuring built-in classifiers that review file writes and command-line operations in real time, automatically blocking high-risk actions such as bulk deletions, data theft, or malicious code execution. Safe operations are executed automatically, while potentially dangerous actions are blocked, prompting the model to adjust its strategy. If the model persists with high-risk behavior, the system will trigger a user permission prompt. Currently, Claude Teams users can access the research preview version, with enterprise and API customers gaining access sequentially.