The Deska blog

Getting Claude Code Running in a Panel

Learn the technical steps for installing Claude Code in Deska to run Anthropic's agent alongside your editor and terminal in a unified canvas.

· 11 min read

Developing software with artificial intelligence has shifted from simple chat interfaces to sophisticated command line agents that can edit files, run tests, and manage git workflows. One of the most powerful tools in this category is Anthropic's CLI agent, but managing its lifecycle alongside your editor often leads to tab fatigue. This guide focuses on installing Claude Code in Deska, allowing you to run this agent as a persistent panel within an infinite canvas workspace. By integrating the agent directly into your environment, you eliminate the friction of switching between a standalone terminal and your code editor.

Understanding CLI Coding Agents

CLI agents represent a departure from the traditional web interface. Instead of copying and pasting snippets, these tools operate directly on your local file system. They are designed to understand the context of a project by indexing files and executing shell commands to verify their own work. This requires a robust terminal environment that can handle long running processes and interactive prompts.

Traditional terminal emulators provide the necessary shell access, yet they often feel disconnected from the visual representation of the project. When you use an agent like Claude Code, it frequently references multiple files or requires you to check a browser preview of a frontend change. Keeping these elements in separate windows can disrupt the flow of development.

Technical Requirements for Claude Code

Before starting the installation process, ensure your local environment is prepared. Claude Code is a Node.js based application, meaning it relies on a modern JavaScript runtime. You will need the following components:

  • Node.js version 18 or higher.
  • A valid Anthropic API key.
  • A compatible shell like zsh or bash.
  • Sufficient permissions to install global npm packages.

Once these basics are covered, you can proceed to integrate the tool into a workspace that supports side by side execution of agents and editors.

Steps for Installing Claude Code in Deska

Deska provides a specialized environment for running these types of tools. Rather than opening a standard OS terminal, you can create a dedicated terminal panel that lives on an infinite canvas. This setup is particularly useful for agents because you can place the terminal directly next to the files the agent is modifying.

  1. Open your workspace and navigate to an empty area on the canvas.
  2. Use the Ask Deska assistant or the command palette to spawn a new terminal panel.
  3. In the terminal panel, run the command npm install -g @anthropic-ai/claude-code to fetch the package.
  4. Once the installation completes, initialize the agent by typing claude in the same panel.
  5. Follow the authentication prompts to link your Anthropic account.

The benefit of this approach is that the session remains active within the panel. Because Deska is local-first, the authentication tokens and session history stay on your machine rather than being synced to a third party cloud.

Comparing Claude Code with Other Agent Workflows

There are several ways to interact with AI agents. Some developers prefer integrated development environment extensions, while others prefer standalone CLI tools. Each approach has trade-offs regarding context awareness and resource usage.

FeatureClaude Code (CLI)IDE ExtensionsDeska Workspace
InterfaceTerminal basedSidebar or InlineInfinite Canvas
ContextLocal file systemActive editor tabMulti panel view
PerformanceHigh (native)Varies by IDENative terminal speed
FlexibilityScriptableFixed UICustomizable layout

While IDE extensions are excellent for quick edits, they can sometimes feel cramped. Standalone CLI tools provide more power but lack visual context. Running agents in panels within a canvas workspace attempts to bridge this gap by providing a visual layout that does not sacrifice the power of the command line.

Managing Multiple Agents Side by Side

One of the unique advantages of using a canvas based system is the ability to run multiple coding agents simultaneously. You might use Claude Code for heavy architectural changes while running another tool like OpenCode for documentation or Codex CLI for quick shell script generation.

To manage this effectively, you can organize your canvas into zones. For example, place your Monaco code editor in the center, a Claude Code terminal on the right, and a browser widget on the left to monitor live changes. This spatial organization helps reduce the cognitive load of tracking what each agent is doing.

If you are working remotely, you can use the mobile app to monitor these long running agent tasks. Since the devices pair directly through a secure relay without exposing ports, you can check the progress of a complex refactor while away from your desk.

Best Practices for Agent Terminals

When running CLI agents in a persistent workspace, keep these tips in mind to maintain a clean environment:

  • Use clear names for your panels to distinguish between different agent sessions.
  • Monitor your API usage, especially if you are using your own keys under a pricing plan that requires it.
  • Leverage the canvas zoom feature to see the "big picture" of your project architecture alongside the agent's granular output.
  • Regularly update the agent package via npm to ensure you have the latest features and security patches.

FAQ

How do I update Claude Code in a terminal panel?

To update the agent, you simply run the command npm install -g @anthropic-ai/claude-code again within any terminal panel. Since the installation is global to your machine, the updated version will be available across all panels and workspaces immediately.

Can I run Claude Code without an internet connection?

Claude Code requires an active internet connection to communicate with Anthropic's models. While your code and files stay local-first on your machine, the reasoning capabilities of the agent are powered by cloud inference. If you lose connection, the agent will typically pause and wait for a reconnect.

Does Deska store my Anthropic API keys?

If you are using the BYOK model, your API keys are stored locally on your device. Deska does not send these keys to its own servers. The workspace is designed to act as a secure wrapper for your existing tools and credentials, maintaining the privacy of your development environment.

Getting Started with Your Own Workspace

Setting up a productive environment for AI assisted development does not have to be complicated. By combining the power of CLI agents with a flexible, canvas based layout, you can create a workflow that fits your specific needs. Whether you are refactoring a legacy codebase or starting a new project, having your tools organized spatially can significantly improve your focus.

To begin building your own setup with Claude Code and other developer tools, you can download the app for Mac, Windows, or Linux. The workspace itself is free to use, allowing you to arrange your terminals, editors, and agents exactly how you want them.

💡 Ideas+🐛 BugsSuggest a feature or report a bug