Anthropic defaults Claude Code to auto mode on August 14
In brief
- Anthropic enables auto mode by default in Claude Code on August 14
- Risk classifier executes low-risk actions without user interruption
- Users previously approved ~93% of prompts under manual system
- Feature available across Anthropic API, AWS Bedrock, Google Vertex
- Administrators can revert to manual approval mode
How auto mode works
Auto mode occupies the middle ground between full autonomy and constant interruption. Rather than asking permission for every action, the system uses a classifier model to evaluate risk before executing anything. This two-stage approach begins with a fast single-token filter that screens incoming requests. If the filter flags something as risky, a chain-of-thought mechanism evaluates the action more carefully before proceeding or halting.
Low-risk actions — reading a file, running a standard build command, or similar routine operations — proceed without interruption. The agent can chain together multi-step operations without stopping to ask permission at each stage. This reduces context switching and keeps developers focused.
Deployment and compatibility
Anthropic first launched auto mode as a research preview on March 24-25, 2026. Since then, the company has refined the risk classifier and expanded support. Model compatibility extends to Sonnet 4.6 and Opus 4.6+. The feature works across multiple deployment environments, including the Anthropic API directly, AWS Bedrock, and Google Vertex.
For teams that prefer the old workflow, there's an escape hatch. Administrators who prefer the old manual approval system can override the default, reverting to per-action approval if their security posture demands it. The rollout gives organizations flexibility while pushing the default toward efficiency.
