Anthropic is also changing the way developers interact with Claude Code as it prepares to make Auto mode the default setting for Pro, Max and Team users from August 14. The move is expected to reduce the number of times developers have to manually approve actions while Claude is coding tasks, giving the AI more freedom to continue through multi-step workflows on its own.
Claude Code is an AI-based coding assistant that can interact directly with a developer's coding environment. Before now, users could be asked to approve individual actions before the system started and approve each action. While the prompts would provide a layer of control, they can also interrupt longer development activities and require developers to be involved in the development process more time-consuming.
With Auto mode as the default, Claude Code will still be able to do anything without permission. Anthropic says the system will still be protected and stop if an action that is irreversible, destructive or outside the user’s working environment occurs. So the change is not about removing human oversight entirely, but rather to reduce unnecessary interruptions during routine coding.
Claude Code to Work More Independently.
Anthropic initially introduced Auto mode as an experimental feature in March. The idea was to make Claude Code feel more free to work with more freedom but with safety mechanisms in place for risky behavior. Anthropic has now decided to make that feature the core experience for several paid user groups.
For developers, the biggest change may be the overall flow of a coding session. Instead of Claude repeatedly pausing to ask if it can carry out the next operation, the AI can continue working through a task until it reaches an action that requires additional approval or intervention.
This is particularly helpful for big programming assignments with many connected steps. For instance, a developer could ask Claude Code to investigate a bug, modify several files, run tests and make the necessary corrections. In Auto mode, the AI can potentially do so without repeatedly interrupting the developer.
At the same time, the ability to work more independently raises the question of how AI coding agents should balance productivity with safety. A more advanced AI can have more control over its development environment, so mistakes could potentially have far-reaching consequences. Anthropic's approach is therefore based upon allowing routine actions and keeping intervention points for operations that are considered very risky.
Anthropic asserts Auto mode is better.
Anthropic also shared its results of Auto mode testing. The feature was tested with 1,053 paid users to see how effectively it could identify potentially harmful actions.
The company said Auto mode detected 89 percent of harmful actions during the testing. Anthropic compared this to human reviewers, who it said identified 13.6 percent of such actions. These numbers are certainly a key part of Anthropic’s argument for making Auto mode default, but they need to be taken into account when considering the company’s own testing methodology and not as a general measure of the safety of AI.
Anthropic has also highlighted productivity gains. Teams and Enterprise users who adopted Auto mode shipped around 25 percent more pull requests. Pull requests are an important part of collaborative software development because they allow developers to review and integrate code changes into shared projects.
If productivity improvement is consistent across teams and development environments, more autonomous coding will be an increasingly important part of software engineering workflows.
Companies are already using Auto Mode.
Anthropic said several companies, including Adobe, Nuro, Gusto and Garner Health, already have Auto mode as their production default. This shows that the company views autonomous operation not simply as an experimental capability but as an important part of the future of Claude Code.
Boris Cherny, Head of Claude Code, has also made a positive comment on the feature. Cherny wrote that he and the Claude Code team have been using Auto mode exclusively for months and can't imagine returning to permission prompts.
In this way, the move is indicative of a wider change in AI development tools. AI systems are no longer just assistants that offer code suggestions but are being designed to be agents that plan tasks, edit files, execute commands, test changes and iterate on their own.
What Developers Should Expect.
For Pro, Max and Team users, the change from August 14 means Auto mode will become the default Claude Code experience. Developers who prefer to approve individual actions may still have options to keep a more controlled workflow depending on settings and permissions in their environment.
The difference is that users will no longer have to consider every routine action as a separate approval decision. Claude Code will be expected to handle more of the workflow independently and ask for human intervention only when it encounters potentially dangerous or out-of-scope operations.
Anthropic's decision is another step toward more autonomous AI coding agents. The success of this approach will ultimately depend on whether developers can gain meaningful productivity benefits without sacrificing control over their code and development environments. As AI systems become capable of handling increasingly complex programming tasks, the balance between autonomy, safety and human oversight is likely to remain one of the industry's biggest challenges.
