I Hired 131 AI Employees Today. Here's How.

Today I hired 131 employees.

They're all AI. They work for free. None of them can quit. I feel slightly powerful.

Here's what happened: I've been using Claude Code (Anthropic's AI coding assistant) for a few months to build and debug things on Vibed Lab. It's weirdly good at knowing when to ask questions and when to just ship code. But at some point I realized the real move was delegation. Claude could spin up specialized AI workers—called subagents—for specific jobs. Someone on GitHub had already assembled 131 of them. This is how I installed all of them in about 10 minutes, and why you should too—even if you've never opened a terminal in your life.

Wait. What's a Subagent?

Think of Claude as a general manager. Smart, capable, handles most things. But one person doing accounting + legal + UI design + server infrastructure is going to do all of those... adequately. Not great.

A subagent is a specialist employee the manager delegates work to.

So you end up with 131 specialists available in an open-source repository. Each one has a specific job, specific tools, specific instructions baked in. You don't write any of that yourself. You just install them and tell the general manager which one to call.

The Repository I Found

It's called awesome-claude-code-subagents, maintained by VoltAgent on GitHub. Right now it's 131 agents across 10 categories:

• Core Development: backend, frontend, fullstack, mobile
• Language Specialists: Python, TypeScript, Rust, Go, React, Vue, Node, Java, C++, and more (26 agents total)
• Infrastructure: Docker, Kubernetes, AWS, Terraform, CI/CD
• Quality & Security: code-reviewer, security-auditor, penetration-tester, debugger
• Data & AI: data-analyst, machine-learning-engineer, data-scientist, prompt-engineer
• Business: product-manager, ux-researcher, legal-advisor, tech-writer, seo-specialist
• Workflow & Coordination: multi-agent-coordinator, deployment-manager

The setup is straightforward: fork the repo (make your own copy on GitHub), clone it to your computer, copy the agent files to where Claude Code looks for them. Total time: about 10 minutes. Here's exactly how.

Installing All 131 (The Actual Steps)

Step 1: Fork the Repository

Go to github.com/VoltAgent/awesome-claude-code-subagents, click the Fork button in the top right.

Why fork instead of just cloning directly? Forking creates your own copy on GitHub. You can track updates, customize agents later, push your own improvements back if you want. It takes 10 seconds and gives you control.

Step 2: Clone It to Your Machine

Open a terminal (Mac: Terminal app, Windows: Git Bash or PowerShell). Run this:

gh repo clone YOUR-USERNAME/awesome-claude-code-subagents

Replace YOUR-USERNAME with your actual GitHub username.

If you don't have the GitHub CLI (gh) installed yet, grab it from cli.github.com. Takes two minutes. Alternatively, you can clone with git clone https://github.com/YOUR-USERNAME/awesome-claude-code-subagents if you already have Git.

Step 3: Install All 131 Agents Globally

Run this one command. It copies every agent file to ~/.claude/agents/ (the folder Claude Code checks):

mkdir -p ~/.claude/agents && \
find ./awesome-claude-code-subagents/categories \
  -name "*.md" ! -name "README.md" \
  -exec cp {} ~/.claude/agents/ \;

What this does in plain English: First, it creates the ~/.claude/agents/ folder if it doesn't exist (the -p flag means "don't error if it already exists"). Then it searches the categories folder for every .md file (these are plain text files that hold each agent's instructions) except for README.md (which is just the project's help document), and copies each one to that agents folder. That's it. No installation scripts, no setup wizards. Just file copying.

Step 4: Verify

Run this to count how many agents got installed:

ls ~/.claude/agents/ | wc -l

Should output 131. If yes, the agents are installed. But installed isn't the same as useful. Here's the part that actually made this worth doing.

The Part That Actually Matters: The Approval Step

After installation, I configured how Claude uses the agents. The default is fine—Claude picks an agent and delegates automatically. But I wanted to stay in the loop. So I set up a 3-step workflow.

Step 1: Claude analyzes your request and picks the right agent(s).

Example: "review this Python code for security issues" → Claude checks the catalog → picks python-pro + security-auditor.

Step 2: Claude asks for approval before doing anything.

"I'm planning to use: — python-pro → analyze code structure — security-auditor → check for vulnerabilities Should I proceed?"

You can say yes, swap an agent, or add one. You're in control.

Step 3: Report comes back in plain language.

No jargon. Each agent explains what it did, what it found, what changed. Like a briefing from your team, not a wall of cryptic error text.

This matters because AI agents can do a lot of work quickly. An approval gate means nothing gets touched without you knowing what's about to happen and why. Setting this up takes about 2 minutes of conversation with Claude Code. Then it persists across all your future sessions.

Do You Actually Need 131 Agents?

Probably not all of them.

I'm a pharmacist who codes as a hobby. I have no use for blockchain-developer or quant-analyst. But I regularly use debugger, security-auditor, python-pro, and technical-writer.

The value isn't having 131. It's not having to decide which ones you need in advance. When you hit a problem you've never seen—and you will—the right specialist is already installed.

It's like a hospital stocking medications it rarely uses. You hope you never need the obscure ones. But when you do, you really, really do.

Here's the honest take: if you use Claude Code at all, this setup takes 10 minutes and costs nothing. Install the agents.

Quick Reference: Which Agent for What

The full list of 131 is in the repository README.

I spent more time writing this post than I did installing the agents. That's probably the best endorsement I can give.

2026.03.07