lockstep
5 min readNaman Jain

Claude Code managed settings: an admin guide

Enforce Claude Code policy across an organization: managed-settings.json paths, delivery options, keys worth setting, a managed CLAUDE.md, and verification.

A security lead asks a simple question: can we guarantee that no engineer's Claude Code reads .env files? With CLAUDE.md, the honest answer is "usually." With managed settings, it is "yes, on every machine that has the policy."

Managed settings are how an organization enforces Claude Code configuration. This guide covers where they live, how to deliver them, the keys worth setting, and how to confirm they took effect, from Anthropic's managed settings, settings, and memory documentation, checked on 16 November 2026.

What managed settings do

Managed settings sit at the top of Claude Code's settings stack. Per the documentation, no user, project, local, or --settings value overrides them, apart from a few security-sensitive keys where a stricter value from a lower level still counts. They use the same JSON shape as any settings.json, so anything you'd put in a project's .claude/settings.json can be enforced company-wide.

The full order, highest first, is explained in Claude Code settings.json.

The quickest start: one file

Anthropic's own first example blocks two kinds of file reads, turns off bypass mode, and makes managed settings the only source of permission rules:

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

Save it as managed-settings.json in the system directory:

OSPath
macOS/Library/Application Support/ClaudeCode/managed-settings.json
Linux and WSL/etc/claude-code/managed-settings.json
WindowsC:\Program Files\ClaudeCode\managed-settings.json

Then run /status on one machine. The Setting sources line should show Enterprise managed settings. Roll out to the rest of the fleet after that.

Delivery options

MechanismHow it reaches machinesUse when
Managed settings filemanaged-settings.json placed with your existing tooling, plus optional drop-ins in managed-settings.d/Servers, containers, CI, or a fleet you already script
MDM policyA plist on macOS, a registry value on WindowsDevices are under MDM
Server-managed settingsJSON in the claude.ai admin console; needs an Owner roleTeams or Enterprise organizations without full MDM coverage

The managed-settings.d/ drop-in directory lets several teams each own a file that merges into one policy. A managed-mcp.json in the same system directory controls MCP servers.

Keys worth setting

GoalKey
Block reading secrets, pushing to main, or other actionspermissions.deny
Stop developers adding their own permission rulesallowManagedPermissionRulesOnly
Turn off bypass permissions modepermissions.disableBypassPermissionsMode
Only run hooks the organization approvedallowManagedHooksOnly
Pre-install the team's pluginsextraKnownMarketplaces with enabledPlugins
Block marketplace sourcesblockedMarketplaces
Limit which models can be usedavailableModels and deniedModels
Require a login method or organizationforceLoginMethod, forceLoginOrgUUID

Each key's entry in the settings reference says whether a managed source can set it. Plugin rollout is covered in Claude Code plugins and marketplaces. Rule syntax is in Claude Code permissions. To enforce the sandbox too, see Claude Code sandbox.

A managed CLAUDE.md

Settings enforce; CLAUDE.md guides. For company guidance every session should see, deploy a CLAUDE.md to the same system directory, or put its content in the claudeMd key:

{
  "claudeMd": "Never commit secrets. Every public API change needs a compatibility note and a test."
}

Developers can exclude other CLAUDE.md files with claudeMdExcludes, but not the managed one. Anthropic's memory documentation is explicit about the split: managed settings for technical enforcement such as blocking tools, a managed CLAUDE.md for behavioral guidance such as code style and data-handling reminders.

Keep it short. A long company-wide file is followed less reliably and crowds out each repository's own setup. For what belongs in a CLAUDE.md, see CLAUDE.md examples.

When a policy doesn't apply

  1. Run /status and check Setting sources.
  2. Run claude doctor; it records sources that exist but couldn't be read, such as a file the OS denied access to.
  3. Check the path for the OS, and that the JSON is valid.
  4. For MDM, check the plist domain or registry key in Anthropic's delivery reference.

What managed settings can't enforce

Managed settings are good at rules you can state as a pattern: this path, this command, this model. They can't hold a decision like "the mobile team stopped supporting iOS 15 last week" or "invoices round per line, not per total, after the Finance review." Those change, have owners, and are made outside any repository or policy file.

Where Lockstep fits

Lockstep covers what policy files don't. Standards & Skills publishes versioned engineering standards and skills and rolls them out by team, repository, path, or task type, with a pilot, expansion, rollback, and time-boxed exceptions per requirement. An adoption view shows which version reached which checkout. For decisions, Lockstep captures them from PRDs, Slack, and pull requests, makes each binding when its owner approves it, and briefs every Claude Code session on the ones that apply.

It works alongside managed settings and supports Claude Code today. See how it works.

Sources checked 16 November 2026

For the rollout plan around these settings, see how to roll out Claude Code to an engineering team. For hooks you might allow, see Claude Code hooks: 6 examples. For the full enterprise picture, see Claude Code for enterprise.

Frequently asked questions

What are Claude Code managed settings?
Settings an organization deploys to every developer's machine. Claude Code applies them above every other level, so no user, project, local, or --settings value overrides them, apart from a few security-sensitive exceptions where a stricter lower value still counts.
Where does managed-settings.json go?
On macOS, /Library/Application Support/ClaudeCode/managed-settings.json. On Linux and WSL, /etc/claude-code/managed-settings.json. On Windows, C:\Program Files\ClaudeCode\managed-settings.json. Drop-in files can also go in a managed-settings.d/ directory beside it.
How do I check that managed settings applied?
Run /status inside Claude Code on one machine. The Setting sources line should show Enterprise managed settings, with the source type. If it's missing, /status and claude doctor record why the policy couldn't be read.
Can I deploy a company-wide CLAUDE.md?
Yes. Place a CLAUDE.md in the same system directory, or put its content in the claudeMd key of managed settings. Developers can't exclude it. Use it for guidance and managed settings for enforcement, because CLAUDE.md is context, not configuration.
Should I use server-managed settings or MDM?
Server-managed settings from the claude.ai admin console suit Teams and Enterprise organizations whose devices aren't all under MDM. If your fleet is already managed, an MDM policy or a managed-settings.json file works and needs no console access.

Keep reading