Claude Code auto mode, Anthropic’s latest update to its developer-focused AI coding tool, gives the model the ability to decide which actions it can take without waiting for user approval. The feature is currently available as a research preview, meaning it is open for testing but is not yet a finished product.
How Claude Code Auto Mode Works
Auto mode uses an artificial intelligence safety layer that reviews each action before it executes. The layer checks for behavior the user did not request and scans for prompt injection attacks — a method where malicious instructions are hidden inside content the AI is processing, causing it to take unintended actions. Safe actions proceed automatically, while flagged actions are blocked.
Extension of an Existing Command
The feature builds directly on Claude Code’s existing “dangerously-skip-permissions” command, which previously handed all decision-making to the AI with no safety filter. Auto mode adds a protective layer on top of that command. However, Anthropic has not detailed the specific criteria its safety layer uses to distinguish safe actions from risky ones.
This lack of transparency is something developers will likely want to understand before adopting the feature at scale. Anthropic has not yet responded to requests for further clarification on this point.
Context: A Broader Industry Shift
The update reflects a wider trend in developer tools, as companies design AI systems to act with less human oversight. Competitors including GitHub and OpenAI have released autonomous coding tools that execute tasks on a developer’s behalf. Anthropic’s approach goes a step further by shifting the decision of when to seek permission from the user to the AI itself.
The challenge across the industry remains consistent: too many approval steps slow development down, while too few create unpredictable and potentially risky systems. Auto mode represents Anthropic’s current attempt to balance those competing demands.
Recent Claude Code Releases and Rollout Details
Auto mode follows two recent additions to the Claude Code platform. Anthropic launched Claude Code Review, an automatic code reviewer designed to catch bugs before they reach the codebase. The company also launched Dispatch for Cowork, which allows users to send tasks to AI agents to handle work on their behalf.
Auto mode will roll out to Enterprise and API users in the coming days. Anthropic said the feature currently works only with Claude Sonnet 4.6 and Opus 4.6. The company recommends using auto mode in isolated environments — sandboxed setups kept separate from production systems — to limit potential damage if something goes wrong.
“Auto mode uses AI safeguards to review each action before it runs, checking for risky behavior the user didn’t request and for signs of prompt injection.”
Anthropic, Company Statement
The security implications of autonomous AI coding tools remain an active area of concern for enterprise developers. Anthropic’s recommendation to use isolated environments signals that the company itself views the feature as carrying meaningful risk at this stage of development.