Anthropic just dropped a new feature called “Auto Mode,” and it’s designed to fix one of the biggest annoyances in AI-assisted coding. The company also revealed that Claude Code has hit $750 million in annualized revenue, with enterprise subscriptions driving a massive surge in growth. This update bridges the gap between manual approvals and unchecked automation.
Why The Shift To Auto Mode Matters
Before Auto Mode, you faced a binary choice: approve every single file change or disable safety checks entirely. That constant back-and-forth was a real barrier to adoption. Now, Anthropic’s introduced a smarter middle ground where the AI handles safe tasks automatically, but it hits a safety wall when it comes to destructive actions.
- Automatic Approvals: Routine operations, like writing files or running tests, get the green light without interruptions.
- Safety First: Anthropic won’t let Claude wipe directories, exfiltrate sensitive data, or execute malicious code by default.
- Trusted Contexts: The AI focuses on your local dev directory and configured Git remotes, ignoring anything else unless you explicitly allow it.
Enterprise Growth & Adoption Metrics
It’s worth noting Anthropic released these numbers alongside the feature update. Claude Code’s annualized revenue has skyrocketed, with enterprise customers now contributing significantly more than individual users. This isn’t just a toy for hobbyists anymore; it’s becoming a serious enterprise tool. The tool has already surpassed 20 million GitHub commits, with 90% of that output landing in low-star repositories, which suggests a large volume of hobbyist use, but it’s the enterprise numbers that are driving the growth.
Implementing The New Safety Features
You’ll notice that Anthropic is clear about risks. They don’t claim this is a sandbox, so they recommend running Auto Mode in containers. The classifier isn’t perfect, so you still need to review code—especially after long autonomous runs. This approach lets you scale up without sacrificing control, which is exactly what enterprise teams need to feel comfortable moving away from manual coding entirely.
