Claude Code Enterprise Administration: Setup, Authentication, and MCP Control

Why Enterprise Claude Code Matters

Claude Code v2.1.149 is not just a developer toy anymore — it is a full-stack AI coding agent that reads, writes, and executes code. For enterprises, that means you need centralized control over:

• Who can authenticate and which models they use
• What servers (MCP) Claude can connect to
• Which commands are allowed or denied
• How updates roll out across your fleet

This guide covers the three pillars of enterprise Claude Code administration.

Installation at Scale

Recommended: Native Installer

The native installer auto-updates in the background and supports all platforms:

# macOS, Linux, WSL
curl -fsSL https://claude.ai/install.sh | bash

# Windows PowerShell
irm https://claude.ai/install.ps1 | iex

# Windows CMD
curl -fsSL https://claude.ai/install.cmd -o install.cmd && install.cmd && del install.cmd

Linux Package Managers (Fleet Deployments)

For managed infrastructure, use signed repositories:

Debian/Ubuntu (apt):

sudo install -d -m 0755 /etc/apt/keyrings
sudo curl -fsSL https://downloads.claude.ai/keys/claude-code.asc \
  -o /etc/apt/keyrings/claude-code.asc
echo "deb [signed-by=/etc/apt/keyrings/claude-code.asc] https://downloads.claude.ai/claude-code/apt/stable stable main" \
  | sudo tee /etc/apt/sources.list.d/claude-code.list
sudo apt update && sudo apt install claude-code

Fedora/RHEL (dnf):

sudo tee /etc/yum.repos.d/claude-code.repo <<'EOF'
[claude-code]
name=Claude Code
baseurl=https://downloads.claude.ai/claude-code/rpm/stable
enabled=1
gpgcheck=1
gpgkey=https://downloads.claude.ai/keys/claude-code.asc
EOF
sudo dnf install claude-code

Alpine (apk):

wget -O /etc/apk/keys/claude-code.rsa.pub \
  https://downloads.claude.ai/keys/claude-code.rsa.pub
echo "https://downloads.claude.ai/claude-code/apk/stable" >> /etc/apk/repositories
apk add claude-code

Version Pinning and Release Channels

Control update cadence with the autoUpdatesChannel setting:

• "latest" — new features immediately (default)
• "stable" — approximately one week behind, skips regressions

{
  "autoUpdatesChannel": "stable",
  "minimumVersion": "2.1.100"
}

To disable auto-updates entirely (for fleet-managed deployments):

{
  "env": {
    "DISABLE_AUTOUPDATER": "1"
  }
}

Binary Integrity Verification

Every release publishes a GPG-signed manifest.json:

# Import the Anthropic signing key
curl -fsSL https://downloads.claude.ai/keys/claude-code.asc | gpg --import

# Download and verify
VERSION=2.1.149
REPO=https://downloads.claude.ai/claude-code-releases
curl -fsSLO "$REPO/$VERSION/manifest.json"
curl -fsSLO "$REPO/$VERSION/manifest.json.sig"
gpg --verify manifest.json.sig manifest.json

Expected fingerprint: 31DD DE24 DDFA B679 F42D 7BD2 BAA9 29FF 1A7E CACE

Authentication Options

Authentication Precedence

Claude Code evaluates credentials in this order:

1. Cloud provider (CLAUDE_CODE_USE_BEDROCK, _VERTEX, _FOUNDRY)
2. ANTHROPIC_AUTH_TOKEN — Bearer token for LLM gateways
3. ANTHROPIC_API_KEY — Direct API key
4. apiKeyHelper — Dynamic credential script
5. CLAUDE_CODE_OAUTH_TOKEN — Long-lived OAuth for CI/CD
6. Subscription OAuth — Interactive /login (default)

Team Authentication Methods

CI/CD Authentication

Generate a long-lived OAuth token for pipelines:

claude setup-token
# Follow OAuth flow, copy the token
export CLAUDE_CODE_OAUTH_TOKEN=your-token

This token is scoped to inference only and expires after one year.

Server-Managed Settings

For organizations without MDM infrastructure, server-managed settings deliver configuration from Anthropic’s servers to every authenticated client.

How It Works

1. Admin configures settings in Claude.ai → Admin Settings → Claude Code → Managed settings
2. Claude Code fetches settings at startup and polls hourly
3. Settings apply at the highest precedence tier (cannot be overridden)

Example: Enforce Permission Deny List

{
  "permissions": {
    "deny": [
      "Bash(curl *)",
      "Read(./.env)",
      "Read(./.env.*)",
      "Read(./secrets/**)"
    ],
    "disableBypassPermissionsMode": "disable"
  },
  "allowManagedPermissionRulesOnly": true
}

Example: Organization-Wide Hooks

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "Edit|Write",
        "hooks": [
          { "type": "command", "command": "/usr/local/bin/audit-edit.sh" }
        ]
      }
    ]
  }
}

Fail-Closed Startup

For high-security environments where an unenforced window is unacceptable:

{
  "forceRemoteSettingsRefresh": true
}

The CLI blocks at startup until settings are fetched. If the fetch fails, Claude Code exits rather than proceeding without policy.

MCP Server Control

MCP (Model Context Protocol) servers extend Claude Code with external tools. In enterprise settings, you need to control which servers are allowed.

Control Patterns

Disable MCP Entirely

Deploy managed-mcp.json with an empty map:

{
  "mcpServers": {}
}

File locations:

• macOS: /Library/Application Support/ClaudeCode/managed-mcp.json
• Linux/WSL: /etc/claude-code/managed-mcp.json
• Windows: C:\Program Files\ClaudeCode\managed-mcp.json

Deploy a Fixed Server Set

{
  "mcpServers": {
    "github": {
      "type": "http",
      "url": "https://api.githubcopilot.com/mcp/"
    },
    "sentry": {
      "type": "http",
      "url": "https://mcp.sentry.dev/mcp"
    },
    "company-internal": {
      "type": "stdio",
      "command": "/usr/local/bin/company-mcp-server",
      "args": ["--config", "/etc/company/mcp-config.json"],
      "env": {
        "COMPANY_API_URL": "https://internal.example.com"
      }
    }
  }
}

Security: Never store API keys in managed-mcp.json — use ${VAR} expansion or OAuth per-user authentication.

Allowlist and Denylist

For policy-based control without exclusive file deployment:

{
  "allowManagedMcpServersOnly": true,
  "allowedMcpServers": [
    { "serverUrl": "https://api.githubcopilot.com/*" },
    { "serverUrl": "https://*.internal.example.com/*" },
    { "serverCommand": ["npx", "-y", "@modelcontextprotocol/server-filesystem", "."] }
  ],
  "deniedMcpServers": [
    { "serverCommand": ["npx", "-y", "unapproved-package"] },
    { "serverUrl": "https://*.untrusted.example.com/*" }
  ]
}

Matching rules:

• Commands match exactly (every argument, in order)
• URLs support * wildcards anywhere
• Denylists always win — nothing overrides a denylist match
• Setting allowManagedMcpServersOnly: true prevents users from broadening the allowlist

Security Considerations

What Server-Managed Settings Cannot Prevent

On unmanaged devices, users with admin access can:

• Modify the Claude Code binary
• Edit cached settings files
• Use third-party providers to bypass server-managed settings

For stronger enforcement, use endpoint-managed settings via MDM (Jamf, Intune, Group Policy).

Credential Management

Monitoring

Enable OpenTelemetry export with OTEL_LOG_TOOL_DETAILS=1 to track:

• Which MCP servers users connect to
• Which tools are invoked
• Session duration and model usage

Quick Reference: Deployment Checklist

1. Choose installation method — native installer for individuals, package managers for fleets
2. Set release channel — stable for production teams
3. Configure authentication — Teams/Enterprise for most, Foundry/Bedrock for cloud-native
4. Deploy MCP restrictions — allowlist approved servers, denylist known-bad
5. Enable server-managed settings — permission denials, hooks, fail-closed startup
6. Verify with claude doctor — confirms installation, auth, and config health
7. Monitor with OTEL — track adoption and compliance

Related Content

• Claude Code on Microsoft Foundry
• Fix Claude Code Installation on macOS
• Claude Code Bootcamp Workshop
• MCP Book Review: Model Context Protocol with LLMs