Tech • IA • Crypto
What's new in Claude Code
8/10TL;DR
Anthropic has introduced a major set of Cloud Code updates focused on smoother developer experience and increased AI autonomy, enabling more independent, parallel, and remote workflows.
KEY POINTS
Shift Toward Autonomy
The latest updates emphasize allowing Claude to act independently during coding sessions, reducing interruptions. New systems assess whether actions are safe or potentially harmful, letting the AI proceed without user approval in non-destructive scenarios. This marks a broader move toward asynchronous development, where tasks continue without constant human oversight.
Remote Control Across Devices
A new remote control feature enables developers to start sessions on one device and continue them on another, including mobile. Sessions can run while users are away, with notifications sent when input is required. This allows coding workflows to extend beyond the desktop without losing continuity.
Flicker-Free Terminal Interface
A redesigned full-screen terminal introduces a virtualized scrollback system, eliminating flickering caused by large outputs. Only visible content is rendered, keeping memory usage stable even in long sessions. The update also adds clickable elements and navigation shortcuts within the terminal interface.
Revamped Desktop Application
The Cloud Code Desktop app has been rebuilt with improved usability. Features include session grouping by project, visibility into execution plans, inline commenting on plans and code, and integrated file diffs. Developers can also interact directly with repositories and pull requests from within the interface.
Auto Mode Reduces Interruptions
The new auto mode allows Claude to bypass routine permission prompts. It uses two checks—whether an action is destructive and whether it resembles prompt injection—to decide if it can proceed autonomously. If risks are detected, the system either finds alternatives or escalates to the user.
Parallel Development with Work Trees
Native support for work trees enables multiple Claude sessions to operate on the same repository without conflicts. Each session works in an isolated copy of the project, simplifying parallel feature development and eliminating manual directory management.
Auto Memory for Persistent Context
A new auto memory system lets Claude retain useful insights across sessions, such as coding preferences and debugging patterns. These are stored locally in structured memory files and loaded selectively to avoid bloating context, improving continuity without sacrificing performance.
Automated Code Reviews
The ultra review feature introduces multi-agent code analysis for pull requests. Different agents scan for bugs, vulnerabilities, and logical errors, followed by validation to reduce false positives. This process significantly shortens the time needed to identify issues.
Routines Enable Background Workflows
Routines allow developers to schedule or trigger Claude sessions via timers, webhooks, or API calls. These workflows can run entirely in the cloud, independent of a user’s machine. Use cases include automated issue triage, repository monitoring, and event-driven coding tasks.
Agent View for Multi-Session Management
A new agent view consolidates multiple active sessions into a single interface. Developers can monitor progress, identify which sessions need input, and send prompts without switching contexts. This supports large-scale parallel workflows.
Enterprise and Deployment Enhancements
Additional updates target enterprise users, including improved Windows support, deployment tools for AWS and Google Cloud, and native binaries for streamlined installation within governance pipelines.
CONCLUSION
The latest Cloud Code updates reflect a clear push toward autonomous, parallel, and device-agnostic development, positioning AI as a more independent collaborator in modern software workflows.