Hello everyone! Today we’re diving into Claude Code — the new buzzword everyone is talking about.
In this article, I’ll walk through the setup step by step and experiment with the new Agents Swarm feature to see whether it really lives up to the hype.
First, What is Claude Code?
Claude Code is an AI-powered command-line interface (CLI) tool from Anthropic that helps developers directly inside the terminal.
But what if you’re not a CLI person and prefer working inside an IDE?
Good news: there’s a Claude Code extension for VS Code.
Just open Extensions in VSC → search for Claude Code → install. That’s it.
Requirements
To use Claude Code, you need a Claude subscription:
Pro, Max, Teams, Enterprise, or a Claude Console account.
Here is the link for the Pricing: https://claude.com/pricing
I personally created a Claude Pro monthly subscription for around $21.40/month.
Now, let us work on Installing Claude Code
On mac, and linux OS run this command
curl -fsSL https://claude.ai/install.sh | bash
For windows powershell, run the command below
irm https://claude.ai/install.ps1 | iex
I’m using a Mac OS, so I ran the first command. The install takes just a few seconds, and you’ll see a success message when it’s finished.
Then, you need to run this additional command to add Claude to your PATH:
echo ‘export PATH=”$HOME/.local/bin:$PATH”’ >> ~/.zshrc && source ~/.zshrc
Now let us verify that everything works by running command “Claude — — help” that successfully lists the different arguments, options and usage for each one of them
Ok, cool, now it is time to see the Claude Code agents swarm feature in action:
Claude Code New Project
First, Let’s create a project and start Claude:
mkdir claude-code-test-project
cd claude-code-test-project
claude
Claude will ask to select the:
- Theme (light/dark)
- Login method
After selecting your option, you’ll be redirected to a browser window to authorize Claude.
I decided to go with Claude pro subscription account.
First, I tried creating an Anthropic Console account, but I had issues confirming payment.
So I recommend that you set up the Claude subscription BEFORE running the tool.
Once login is successful, you see the message below on the screen:
Close the browser window, and return to the terminal and press Enter.
Then Claude will:
- Display security notes, press enter to continue
- Terminal settings, select recommended one and press enter
- Final security notification about files you work with when using Claude code agent, select yes and hit enter to continue
At this point, you’re ready. Let us explore existing functionality:
Exploring Commands
Claude Code uses special prefixes:
- ! → Bash mode
- / → Claude commands
- @ → File paths
- $ → Background tasks
Type / and you’ll see available commands.
The most interesting ones are:
- /agents
- /chrome
Today, we’re focusing on /agents and building a swarm of collaborating agents.
Creating Specialized AI Agents
For testing, I’ll create two agents with different perspectives:
1. Robert Greene–style Strategic Advisor
This agent analyzes:
- Power dynamics
- Human nature
- Long-term strategy
Second agent
2. Martin Luther King — Moral Counselor
This agent focuses on:
- Justice
- Empathy
- Ethical leadership
- Transforming conflict
These two agents will collaborate together to give advice for questions related to the domains they cover.
Ok, i’ll start creating the strategic advisor agent,
Use the agents command, create a new agent,
You’ll notice you can create a local agent tied to a specific project, or a general agent that works across all projects. For this demo, I’ll choose the Project option.
then write your detailed prompt., and Claude will generate their configuration.
Once done, repeat the same process with a different prompt for the moral agent.
To confirm that both agents are created successfully, use command /agents to list them
You should see the two project agents listed. The strategic advisor and the moral counsel
Now The Moment of Truth — Agent Swarm in action!
Let’s test if both agents collaborate effectively.
I’ll ask a question that falls within the domain of both agents
“What are the top 3 war and peace strategies?
Also tell me which agents you used to answer.”
Claude then activates both agents in parallel!
You’ll see:
- Strategic analysis from the Robert Greene agent
- Moral and justice-based perspectives from the Martin Luther King agent
- An additional section of where both viewpoints converge
And honestly — this part is super cool.
You get stable, grounded perspectives because each agent stays consistent with its persona, instead of a single agent switching viewpoints to please the user 🙂
That was really impressive and easy to set up!
Claude made a breakthrough with these powerful, capable agents — and it definitely put other competitors behind.
That’s it for today — I hope to enjoy this article about Claude Code swarm agents.
Author
Noor Sabahi
Senior AI & Cloud Engineer | AWS Ambassador
#ClaudeCode #AgentsSwarm #GenerativeAI #AgenticAI #AWSAmbassador