What is Claude Code and what can it do?

Claude Code is Anthropic's AI tool designed to autonomously manage development workflows. It can directly control computers to perform tasks like opening files, using web browsers and applications, and running developer tools. Claude Code functions as an active agent capable of reading codebases, reasoning about tasks, writing and editing files, and executing commands from start to finish, with explicit user permission.

How much code is Claude Code expected to write?

Claude Code is projected to author a significant portion of global code in the near future. Currently, Claude Code authors 4% of all global GitHub commits, and this figure is expected to reach 20% by the end of 2026. Some companies, like Spotify, have already shifted to AI-assisted coding, and Anthropic's team ships 10-30 pull requests per day per engineer, all generated by Claude.

With what companies is Anthropic partnering to expand Claude's capabilities?

Anthropic is partnering with companies like Xero to extend Claude's reach beyond code generation. The partnership with Xero, an accounting platform with 4.6 million subscribers, embeds Claude directly into Xero's product and integrates Xero's financial data into Claude.ai. This allows Claude to assist with small business finances and demonstrates a broader adoption pattern where Claude becomes the interface for interacting with professional tools.

How does Claude Code ensure user control and security?

Claude Code always seeks explicit user permission before exploring, scrolling, or clicking to complete a task, maintaining a crucial 'leash' on its enhanced control. This ensures that users retain control over the AI's actions and prevents unauthorized access or modifications. This approach aligns with Anthropic's vision for AI to handle the mechanical aspects of software development while keeping humans in control.

Claude Code Handbook: Master AI Dev & Autonomous Coding with Claude

Anthropic now gives its AI more autonomous control over computers, allowing Claude Code and Cowork tools to execute complex tasks on users' behalf. This advancement transforms AI from a sophisticated assistant into an active agent, capable of managing development workflows, interacting with applications, and even financial platforms with explicit user permission. The move aims to eliminate coding "minutia," freeing human developers and non-technical users to focus on high-level judgment and creative problem-solving. This marks a significant shift in AI-assisted development, heralding a new phase of automation.

Claude Code Gains Autonomous Control

Anthropic is equipping its Claude Code and Cowork AI tools with advanced autonomous capabilities, allowing the AI to directly control user computers to perform tasks. This new feature enables Claude to open files, use web browsers and applications, and run developer tools without manual setup, functioning as a true agent rather than just a text generator, according to The Verge. Initially available as a research preview for Claude Pro and Max subscribers, this functionality is currently limited to macOS.

This update builds upon prior autonomous features introduced in 2024 with the Claude 3.5 Sonnet model. Anthropic emphasizes that Claude will always seek explicit user permission before exploring, scrolling, or clicking to complete a task, maintaining a crucial "leash" on its enhanced control, TechCrunch reports. This capability aligns with Anthropic's vision for AI to handle the mechanical aspects of software development.