Running Claude Code on Two Projects at Once

Managing modern software development often requires hovering between different repositories. You might be updating a frontend application while simultaneously tweaking the backend API it consumes. Traditional terminal workflows can become cluttered when you try running Claude Code on two projects at once. The difficulty lies in maintaining the mental model of both file systems without losing context in a single, narrow window.

The Challenge of Multi-Project Context

When developers work with AI agents like Claude Code, the quality of the output depends heavily on the provided context. If you are working across two distinct codebases, the agent needs access to both sets of files to understand how changes in one affect the other.

Standard terminal environments usually force a choice. You either open two separate terminal windows and manually switch back and forth, or you use a multiplexer like tmux. While these are powerful tools, they often hide information from view. When you are deep into a complex refactor that spans a client-side React app and a server-side Go service, having to remember which tab holds which project state adds cognitive load.

Multi-project search becomes particularly difficult in these scenarios. Searching for a specific function name or API endpoint across two repositories often involves running separate grep commands or having two instances of an IDE open. This fragmentation makes it harder for an AI agent to see the whole picture.

Approaches to Managing Multiple Agents

There are several ways to organize your workspace when you need to interact with multiple projects and AI agents simultaneously.

1. The Side-by-Side Method: This involves splitting your screen vertically or horizontally. It works well for two projects but scales poorly once you add a third project or a browser window for documentation.
2. The Tabbed Workflow: This is common in IDEs. Each project is a different root folder. The downside is that AI agents in one tab may not have awareness of the files in another tab unless explicitly configured.
3. The Virtual Workspace: Using a canvas or a boundless desktop allows you to lay out the internal structures of both projects. You can see the backend logs in one corner and the frontend agent output in another.

Leveraging Canvas Workspaces for AI Agents

A canvas-based approach changes how you interact with tools. Instead of being confined to a grid or a tab, you can place your terminals exactly where they make sense logically. This is where Deska provides a different perspective on workflow organization.

In a canvas environment, you are not limited by the edges of a single window. You can zoom out to see the high level architecture of your multi-project setup and zoom in to specific agent interactions. When you intent to run Claude Code on two projects at once, you can place two terminal panels side by side. Each panel can be initialized in a different directory.

This setup allows for a more fluid interaction. You can watch one agent generate a migration script in your database project while another agent updates the corresponding ORM models in your application project. Because the workspace remains local-first, the speed of access to your files remains high, and your data stays on your machine. You can find more details on how this works in the documentation for data and storage.

Comparison of AI Agent Interfaces

Different tools take different approaches to how they handle multiple concurrent tasks.

Standard terminals are excellent for focused, single-task work. IDEs offer deep integration with language servers but can feel heavy when you only need to run a quick agent command. A workspace like Deska sits in the middle, offering the flexibility of multiple terminals with the visual organization of a design tool.

Synchronizing Work with Ask Deska

One of the hurdles when working on two projects is synchronization. If you change a port in your backend, you must remember to update the frontend environment variables.

Using an assistant like Ask Deska helps bridge this gap. Because the assistant can drive the workspace, it can help you open the necessary panels for both projects simultaneously. You can ask the assistant to check the status of your sessions or run specific commands across the canvas. This reduces the manual labor involved in setting up your multi-project environment every morning.

For developers who need to step away from their main machine, the ability to monitor these long-running agent tasks is crucial. The mobile app allows you to check on the progress of your Claude Code sessions through a secure relay. It pairs directly with your desktop, ensuring that you can see if a build failed or if an agent is waiting for your input without exposing ports to the public internet.

Security and Privacy in Multi-Project Environments

When working on sensitive projects, privacy is a major concern. AI agents often require access to your entire codebase to be effective. Keeping this process local-first ensures that your files are not being uploaded to a third party server for processing or storage.

The use of your own API keys, often referred to as BYOK, gives you control over your costs and data usage. Whether you are using Claude Code, Codex CLI, or OpenCode, managing these coding agents within a secure, local workspace is a best practice for professional development.

Best Practices for Multi-Project Search

To make the most of your dual-project setup, consider these strategies:

• Use clear naming conventions for your workspace panels so you do not confuse the two projects.
• Group related tools together. Keep the backend terminal next to the backend logs and the frontend terminal next to the browser preview.
• Utilize notes within your workspace to keep track of cross-project dependencies or shared API contracts.
• Map global shortcuts to quickly navigate between the different areas of your canvas.

FAQ

How to search code across multiple repositories?

Searching across multiple repositories is most effective when you use a tool that can index several root directories simultaneously. You can either use command line tools like ripgrep across a common parent folder or utilize a workspace that allows you to open multiple file explorer panels side by side for manual inspection.

Can Claude Code edit two projects simultaneously?

Claude Code typically runs within a specific directory context. To edit two projects at the same time, you should run two separate instances of the agent in two different terminal sessions. An infinite canvas workspace is ideal for this because it allows you to monitor both instances without switching tabs.

Is it possible to share context between two AI agents?

Currently, most AI agents operate independently within their own session threads. To share context, you often need to manually copy specific outputs from one agent to the other. Using a central chat assistant that has visibility over the entire workspace can help coordinate the information between these separate agents.

Get Started with Deska

If you are looking for a better way to manage your AI driven development workflow, you can download the application for Mac, Windows, and Linux. The workspace itself is free to use and provides a flexible environment for all your coding agents.

Download Deska