5 simple Claude Code workflows you must have - beginners guide

Key Concepts

• Multi-Agent Cloud Code Setup
• Pull Request Review with Cloud Code
• Claudia UI for Cloud Code
• MCPS (Machine Communication Protocols) Integration with Cloud Code
• Large File Refactoring with Cloud Code

1. Multi-Agent Cloud Code Setup

• Main Idea: Using multiple Cloud Code agents working together on a single project for complex tasks.
• Workflow:
  1. Setup: Create a coms.md file for inter-agent communication and a cloud.md file as the central system prompt for all agents.
  2. Agent Configuration: Assign roles to each Cloud Code instance (e.g., Agent 01, Agent 02, Agent 03).
  3. Communication: Agents communicate through the coms.md file to coordinate tasks and share progress.
  4. Execution: Launch each agent with specific instructions and allow them to collaboratively work towards a common goal.
• Example: Creating an interactive 3D mountain terrain game.
• Key Prompt: "Read both MD files in our project to develop a full understanding of what our goal is as well as how you need to work with other AI agents in this project."
• Challenge: Taming the agents to work together effectively and not against each other.
• Technical Details: Using Grok-4 as the chief prompt engineer to construct rules for the multi-agent system. Switching Cloud Code instances to auto-accept mode (Shift+Tab) and using the Opus model (/model opus).
• Logical Connection: This sets the stage for demonstrating the power of coordinated AI agents and provides a foundation for more complex workflows.

2. Pull Request Review with Cloud Code

• Main Idea: Using Cloud Code to assist in reviewing pull requests, improving efficiency and accuracy.
• Workflow:
  1. Stage 1: Understanding the PR: Use GitHub CLI commands to view the PR and understand its purpose.
  2. Stage 2: Branch Switching: Switch to the branch associated with the pull request.
  3. Stage 3: Manual Testing: Test the changes made by the pull request (e.g., verifying the 1500 character limit).
  4. Stage 4: File Analysis: Analyze the changes in each file affected by the pull request, using Cloud Code to explain the changes.
• Example: Reviewing pull request #1926 for the speaker's AI startup, which implements a 1500 character limit for ideas in the app.
• Key Argument: Combining human expertise with LLM capabilities leads to faster and more thorough pull request reviews, catching bugs that humans or LLMs might miss individually.
• Technical Details: Using a highly optimized prompt (over 90 lines of context engineering) for pull request review, found in the "New Society" resource.
• Notable Quote: "What's better than one pair of eyes? Two pair of eyes, right? And especially if one of them is human and one is LLM."
• Logical Connection: This builds on the previous workflow by demonstrating how Cloud Code can be applied to a specific software development task.

3. Claudia UI for Cloud Code

• Main Idea: Using a UI-based tool called Claudia for managing and running Cloud Code agents, offering an alternative to the terminal.
• Details:
  • Claudia is an open-source project that provides a user-friendly interface for managing multiple Cloud Code agents in parallel.
  • It eliminates the need for using the terminal or an IDE.
  • It requires installing Rust.
  • The GitHub repository contains setup instructions.
• Argument: Claudia provides a more accessible option for users who prefer a UI over the command line.
• Technical Details: The project reached over 8,000 stars on GitHub in three weeks.
• Logical Connection: This offers an alternative method for interacting with Cloud Code, appealing to users who prefer a visual interface.

4. MCPS (Machine Communication Protocols) Integration with Cloud Code

• Main Idea: Using Cloud Code with MCPS to manage tasks and integrate with other applications like Vectal.
• Workflow:
  1. MCP Setup: Configure MCP servers within Cloud Code using the /mcp command.
  2. API Key Generation: Generate an API key for MCP in applications like Vectal (Vectal Pro users or above).
  3. Configuration: Install the remote package (npm install -g remote) and copy the JSON schema to Cloud Code or another MCP client.
• Example: Using Vectal to manage tasks and running Cloud Code on those tasks.
• Argument: Integrating Cloud Code with MCP-enabled applications streamlines task management and improves efficiency.
• Technical Details: Using Vectal's advanced settings to generate an API key for MCP.
• Logical Connection: This expands the possibilities of Cloud Code by integrating it with external applications and task management systems.

5. Large File Refactoring with Cloud Code

• Main Idea: Using Cloud Code to refactor large, complex code files into more modular and maintainable components.
• Core Philosophy: Treating large file refactoring like a surgery on a live patient, emphasizing safety and precision.
• Workflow:
  1. Phase 1: Safety Net:
     • Write tests for 80-90% (or close to 100%) behavior coverage to establish a baseline.
     • Track feature performance and reliability before refactoring.
  2. Phase 2: Surgical Planning:
     • Identify complexity hotspots and prioritize refactoring based on risk.
     • Focus on extracting the safest and lowest-risk blocks of code first (50-150 lines).
  3. Phase 3: Incremental Execution:
     • Extract code in small, manageable steps (e.g., 40-60 lines).
     • Run tests after each extraction to ensure functionality is maintained.
• Key Argument: Proper refactoring is essential for scaling AI startups and maintaining a readable, modular, scalable, and maintainable codebase.
• Notable Quote: "The core philosophy is treating large file refactoring like a surgery on a live patient. One wrong cut can kill the whole system."
• Technical Details: The presenter emphasizes the importance of testing and incremental changes to avoid introducing bugs during refactoring.
• Logical Connection: Addresses a critical aspect of software development often overlooked – the importance of refactoring for long-term maintainability and scalability.

Synthesis/Conclusion

The video presents five practical Cloud Code workflows designed to enhance developer productivity and streamline various software development tasks. These workflows range from multi-agent collaboration and pull request reviews to UI-based management, MCP integration, and large file refactoring. The speaker emphasizes the importance of combining human expertise with the power of AI to achieve optimal results, particularly in complex tasks like refactoring. The key takeaway is that Cloud Code, when used strategically, can significantly improve efficiency, reduce errors, and enable developers to build and scale their projects more effectively. The mention of New Society and Vectal Ultra is a promotional element, but the underlying workflows offer valuable insights for developers looking to leverage Cloud Code in their daily work.