Part of the People Ops AI Brain - ← Back to Hub

┌──────────────────────────────────────────────────────────────────────────────────┐
│              ⚡ CLAUDE CODE FOR PEOPLE TEAMS                                     │
│                    Mental Model                                                  │
├──────────────────────────────────────────────────────────────────────────────────┤
│                                                                                  │
│   THE DIFFERENCE:                                                                │
│   ┌──────────────────┐  ┌──────────────────┐  ┌──────────────────┐               │
│   │ CO-WORK          │  │ CLAUDE CODE      │  │ CODE + n8n + MCP │               │
│   │ • Documents      │  │ • Infrastructure │  │ • Full systems   │               │
│   │ • Artifacts      │  │ • Workflows      │  │ • Autonomous ops │               │
│   │ • Collaboration  │  │ • Code & config  │  │ • Connected AI   │               │
│   │ • You + Claude   │  │ • Claude builds  │  │ • Claude builds, │               │
│   │   on a doc       │  │   for you        │  │   deploys, runs  │               │
│   └──────────────────┘  └──────────────────┘  └──────────────────┘               │
│                                                                                  │
│   THE STACK:                                                                     │
│   ┌─────────────────────────────────────────────────────────────┐                 │
│   │  Claude Code (brain)                                       │                 │
│   │    ↕ MCP connections                                       │                 │
│   │  n8n (orchestration)                                       │                 │
│   │    ↕ API integrations                                      │                 │
│   │  Your systems (Notion, Slack, HRIS, ATS, Google)           │                 │
│   └─────────────────────────────────────────────────────────────┘                 │
│                                                                                  │
│   WHAT THIS UNLOCKS:                                                             │
│   "Build me an automation that..." → Claude Code builds it                       │
│   "Connect this to our HRIS" → Claude Code wires it up                           │
│   "Test it and fix the errors" → Claude Code debugs itself                       │
│   "Deploy it" → Claude Code ships it                                             │
│                                                                                  │
└──────────────────────────────────────────────────────────────────────────────────┘

<aside> ⚡

TL;DR: Claude Code is an agentic coding tool that runs in your terminal. Connected to n8n via MCP, it can build, test, debug, and deploy entire People Ops automation workflows from a natural language conversation. It's one of the most powerful tools in the orchestration stack - but it's part of a broader systems approach, not the whole story. This is the advanced path for people who've built their first automations and want to move from manual workflow building to AI-built infrastructure.

Read time: 20 min | Delegate to: Technical People Ops lead or AI champion | Requires: Comfort with the terminal, n8n instance running, willingness to go deep

</aside>

<aside> ⚠️

This is the advanced guide. If you haven't yet:

• Built your first automation → start with Building Your First Automation
• Set up MCP connections → start with Connecting Your Systems
• Used Claude Co-Work → start with Claude Co-Work for People Teams

This guide assumes you've done all three and want to go further.

</aside>

<aside> 🔥

What's under the hood (April 2026 update): Claude Code now runs on Opus 4.7, which shipped 16 April 2026. What matters for you:

• 1M token context window at standard pricing. No beta, no long-context surcharge. Claude Code can hold your entire project in its head. CLAUDE.md, workflow files, skills, Notion schemas. All of it.
• xhigh effort level. A new tier between high and max. Claude Code now defaults to xhigh for coding, which means more compute per turn and higher quality output. Use /effort to adjust in-session.
• Auto mode (Max plans). A classifier handles your permission prompts. Safe actions run without interrupting you. Risky actions still stop for review. This sits between approving everything and --dangerously-skip-permissions, which is where most People Ops builders actually want to be.
• Agent teams. Multiple Claude Code agents working in parallel, coordinated automatically. Shipped as research preview with Opus 4.6, matured with 4.7.
• Task budgets (beta). Give a long-running agent a token budget and it paces itself. The model sees a running countdown and prioritises accordingly. This is the feature that stops you getting surprise bills on a runaway workflow.
• Computer use in the CLI (research preview). Claude Code can open native apps and click through UIs from the terminal when it needs to verify something only a GUI can show.

One thing to watch. The new tokenizer uses 1 to 1.35 times more tokens than 4.6, depending on content. Same headline price, but real cost per task can land 10 to 35 percent higher. Plan accordingly.

</aside>

Where Claude Code Fits in the Stack

Before we get into the details: Claude Code is a build tool. It's how you construct automation infrastructure quickly. But it sits inside a broader orchestration approach - your reasoning model (Claude), your workflow layer (n8n), your connected systems (MCP), and your governance framework. Claude Code accelerates the building. The architecture decisions still need a human.

If you're looking for the bigger picture on how all these pieces fit together, start with Connecting Your Systems and the orchestration framing in Understanding Agentic AI.

What Is Claude Code?

Claude Code is Anthropic's agentic coding tool. It runs in your terminal. Not the browser. Not a chat window. Your actual terminal. It reads files, writes files, executes commands, runs scripts, calls APIs, and connects to external systems via MCP.

If that sounds like a lot of power for a People Ops person to wield... good. That's the point.

The critical difference from Co-Work:

Co-Work

• You collaborate on documents
• Output is artifacts (text, docs, strategy)
• You work with Claude side by side
• Best for thinking work - handbooks, frameworks, strategies
• Like sitting next to a smart colleague

Claude Code

• You direct the building of systems
• Output is working infrastructure (automations, integrations, workflows)
• Claude builds for you while you supervise
• Best for building work - automations, connections, tooling
• Like having a senior engineer on your team

<aside> 💡

The mental model: Co-Work is your thinking partner. Claude Code is your builder. Co-Work helps you design the compensation philosophy. Claude Code builds the automation that routes comp review approvals through Slack, updates the tracker in Notion, and sends the confirmation email.

</aside>

Why This Matters for People Teams

You've done the work. You've learned to prompt properly, built some skills files, maybe even got an automation or two running in n8n. And then you hit the wall.

Not the "this doesn't work" wall. The "this works but it takes me four hours to build each workflow by clicking through nodes" wall. The value is obvious. The bottleneck is the build time.