AI Briefing
KOSign in

Berthd Releases Open-Source macOS Client for Persistent Remote Coding Agents

·2026.10.07 04:14

Key point

Berthd offers an open-source macOS client that runs Claude Code and Codex on remote dev boxes, allowing agents to continue working even when the local laptop is closed.

1 / 7

Details

Berthd has released an open-source macOS application designed to decouple AI coding agents from local hardware constraints. The tool runs Claude Code, Codex, and other agents on user-controlled development boxes, allowing workflows to persist independently of the local machine's state.

Core Functionality

The application provides a unified interface for managing remote agent sessions:

  • Persistent Execution: Agents continue running on remote boxes even when the user's laptop is closed or disconnected.
  • Worktree Management: Users can initiate tasks via a simple command (⌘N), which creates a git worktree on the selected box and launches the chosen agent within it.
  • Real-Time Monitoring: The interface displays live command steps, code edits as diffs, and interactive prompts from the agent, allowing users to answer questions or approve actions in place.

Interface and Features

Berthd emphasizes a developer-centric workflow with specific tools for comparison and customization:

  • Side-by-Side Comparison: Users can view two worktrees simultaneously in mirrored lanes, comparing chats, diffs, and terminals to evaluate different agent outputs.
  • Artifact Tracking: Published artifacts are listed within the chat interface for easy access.
  • Customization: The app supports 19 theme ports (including Dracula, Rosé Pine, Catppuccin, and Tokyo Night) and allows custom chat backgrounds.

Availability

The macOS client is signed, notarized, and available for both Apple silicon and Intel Macs. The server component, berthd, is installed on remote boxes via a shell script. The project is free and open source.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.