Enterprise Patterns: Teams and Organizations

Claude Code at Scale

Individual Claude Code usage is straightforward. Scaling it across a team of 10, 50, or 500 developers requires standardization, policies, and processes. This lesson covers the patterns that work.

The Configuration Stack

Enterprise Claude Code has four configuration layers:

/etc/claude-code/           ← IT-managed, highest priority
  └── settings.json         ← Enforced policies (can't override)
~/.claude/settings.json     ← Personal defaults
.claude/settings.json       ← Team standards (committed to git)
.claude/settings.local.json ← Personal overrides

/etc/claude-code/ — Managed Policies

IT creates /etc/claude-code/settings.json with enforced policies:

{
  "permissions": {
    "deny": [
      "Bash(curl *)",
      "Bash(wget *)",
      "Bash(nc *)",
      "Bash(ssh *)",
      "Bash(sudo *)",
      "Bash(rm -rf *)",
      "Read(.env*)",
      "Read(*/.pem)",
      "Read(*/.key)"
    ]
  }
}

These denials cannot be overridden by any other settings layer. Even if a developer adds Bash(curl *) to their personal allow list, the managed deny takes precedence.

Team Settings — .claude/settings.json

Committed to git, shared with everyone:

{
  "permissions": {
    "allow": [
      "Read",
      "Write(src/**)",
      "Write(tests/**)",
      "Bash(pnpm *)",
      "Bash(git *)"
    ],
    "deny": [
      "Write(package.json)",
      "Write(.github/**)",
      "Bash(pnpm add *)"
    ]
  },
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "Write(*.ts)",
        "command": "npx prettier --write $file"
      }
    ]
  },
  "env": {
    "ENABLE_LSP_TOOL": "1",
    "MAX_THINKING_TOKENS": "10000"
  }
}

Shared CLAUDE.md

The team CLAUDE.md is the single most important file for standardization. Commit it to git:

# Project: ShipFast API

Stack
Node.js 20 + TypeScript 5.4
Fastify 4.x + Prisma 5.x + PostgreSQL 16

Architecture Rules
[Team-agreed rules that everyone follows]

Coding Conventions
[Team-agreed conventions]

Commands
[Build, test, deploy commands]

Agent Team Rules
[How sub-agents should behave]

For New Team Members
Run pnpm install before starting Claude Code
Set ENABLE_LSP_TOOL=1 in your shell
Read docs/onboarding.md for project context

When anyone on the team starts a Claude session, they get the same instructions.

Custom Endpoints: modelOverrides

Route Claude Code through your infrastructure:

AWS Bedrock

{
  "modelOverrides": {
    "claude-sonnet-4-20250514": {
      "apiKeyEnvVar": "AWS_ACCESS_KEY_ID",
      "baseURL": "https://bedrock-runtime.us-east-1.amazonaws.com",
      "model": "anthropic.claude-sonnet-4-20250514-v1:0"
    }
  }
}

Google Vertex AI

{
  "modelOverrides": {
    "claude-sonnet-4-20250514": {
      "apiKeyEnvVar": "GOOGLE_APPLICATION_CREDENTIALS",
      "baseURL": "https://us-east5-aiplatform.googleapis.com",
      "model": "claude-sonnet-4@20250514"
    }
  }
}

Why Custom Endpoints?

• Compliance: Data stays in your cloud region
• Monitoring: Track usage through your API gateway
• Cost management: Centralized billing
• Access control: Manage who can use which models

Onboarding New Team Members

The Onboarding Checklist

Create .claude/commands/onboard.md:

---
description: "Onboard a new team member to Claude Code"


Welcome to the team! Let me help you get set up with Claude Code.

1. Check Environment:
   - Verify Node.js version: node --version (must be 18+)
   - Verify Claude Code: claude --version
   - Run claude doctor for diagnostics

2. Review Project Context:
   - Read CLAUDE.md for project standards
   - Read .claude/settings.json for permissions
   - List available custom commands: show all .md files in .claude/commands/

3. Verify Setup:
   - Check that LSP is enabled
   - Check that auto-formatting hooks work
   - Run a test: pnpm test should pass

4. Quick Tour:
   - src/services/ — business logic
   - src/routes/ — API endpoints
   - src/repos/ — database queries
   - tests/ — test files

Explain each item as we go through it.

New team member runs:

/onboard

Claude walks them through setup verification and project orientation.

The Documentation Set

What to include in your repo for new Claude Code users:

.claude/
├── settings.json          ← Team permissions and hooks
├── commands/
│   ├── onboard.md         ← Onboarding guide
│   ├── review.md          ← Code review command
│   ├── test.md            ← Test generation command
│   └── deploy.md          ← Deploy checklist
├── agents/
│   ├── security-reviewer.md
│   └── test-writer.md
└── skills/
    └── api-endpoint/
        └── SKILL.md
CLAUDE.md                  ← Project standards
.mcp.json                  ← Team MCP servers

Standardizing Across Multiple Repos

For organizations with many repos:

Template Repository

Create a template repo with standard Claude Code configuration:

company-claude-template/
├── .claude/
│   ├── settings.json      ← Standard permissions
│   ├── commands/           ← Standard commands
│   └── agents/             ← Standard agents
├── CLAUDE.md.template      ← Template with TODOs for customization

When creating a new project, start from this template.

Organization-Wide Commands

Share commands across projects via the global directory:

~/.claude/commands/
├── company-review.md       ← Company code review standards
├── company-security.md     ← Company security checklist
└── company-standup.md      ← Standup report format

Distributed to all developers as part of their setup.

Key Takeaway

Enterprise Claude Code requires four things: managed policies at /etc/claude-code/ for IT-enforced security, shared team configuration in .claude/settings.json committed to git, a comprehensive CLAUDE.md for project standards, and custom endpoints via modelOverrides for compliance. Onboard new members with a /onboard command, standardize across repos with templates, and share common commands globally. The goal is: any developer, any project, same Claude Code experience.