Settings

JSON Schema reference for Claude Code settings

Pick which model answers when Claude asks the advisor tool

Start every session as a named subagent with its prompt, tools, and model

Let Claude send a push notification to your phone when it decides to

Load the claude.ai connectors Claude Code fetches itself alongside a deployed `managed-mcp.json`

Run only the hooks your organization deploys

Make the managed MCP allowlist the only one that applies

Make managed settings the only settings source of permission rules

Replace the default allowlist of channel plugins that can push messages

Limit which URLs HTTP hooks can target

Allowlist which MCP servers people can use

Turn extended thinking off for every session

Generate the API credential with your own command

Let an unanswered question auto-continue after idle time

Customize the attribution Claude Code adds to commits and pull requests

Change or hide the trailer Claude Code adds to commits

Change or hide the attribution line in pull request descriptions

Omit the claude.ai session link from cloud and Remote Control commits

Turn automatic compaction off or on

Set how full the context gets before Claude Code compacts

Connect to a running VS Code or JetBrains IDE automatically from an external terminal

Wait in the open session and continue the task automatically after a claude.ai usage limit resets

Turn off automatic install of the IDE extension from a VS Code terminal

Store auto memory in a directory you choose

Turn auto memory off or on

Add your own allow and deny rules to the auto mode classifier

Send every shell command through the auto mode classifier, even ones a narrow allow rule matches

Follow new output to the bottom in fullscreen rendering

Follow the stable release channel instead of latest

Restrict which models people can pick

Turn off the session recap shown when you come back to the terminal

Refresh expired Bedrock credentials in `.aws` with your own command

Supply Bedrock credentials as JSON from your own command

Render screen-reader friendly output

Block plugin marketplace sources for your organization

Keep Claude's tools off external pages in the desktop Browser pane

Allow channels for your organization

Inject organization-wide CLAUDE.md instructions from managed settings

Skip specific CLAUDE.md files when memory loads

Choose how many days Claude Code keeps transcripts before deleting them

Show your organization's announcements at startup

Choose whether Claude Code delivers messages from your other sessions, shows a notice without delivering them, or refuses them

Choose whether Bash or PowerShell runs the shell commands you type with the `!` prefix

Block specific MCP servers by URL, command, or name

Set an age limit in days for Claude Desktop and Cowork transcripts

Set how long Claude Code waits for Remote Control or an SDK host to answer a forwarded dialog before it cancels the dialog

Choose whether Claude's proposed file changes open in the VS Code or JetBrains diff viewer or stay in the terminal

Turn off background agents and agent view

Turn off hooks, a custom status line, and a custom `@` file suggestion command at once

Deprecated; use `enableArtifact` to turn the Artifact tool off

Remove auto mode from the permission mode cycle

Limit the desktop Browser pane to localhost for people and Claude

Turn off the skills and workflows included with Claude Code

Turn off claude.ai connectors so Claude Code doesn't fetch them

Block plugins that install by running a marketplace-declared command

Stop Claude Code from registering the `claude-cli://` handler

Turn off Desktop Code sessions that run on the device, leaving SSH to other hosts and cloud

Block Claude's tools in the desktop iOS Simulator pane

Turn off Remote Control everywhere it can start

Reject the CLI flags that sideload plugins, subagents, and MCP servers

Stop skills and custom commands from running inline shell

Turn dynamic workflows off for everyone; use `enableWorkflows` for yourself

Reject specific servers from a project's `.mcp.json`

Use vim key bindings in the input prompt

Set a default effort level for models without a saved level of their own

Turn off `:shortcode:` emoji suggestions and replacement in the prompt input

Approve every server in project `.mcp.json` files without a prompt

Turn the Artifact tool off with a `false` in any file; no file can turn it back on

Turn dynamic workflows on or off against your plan's default

Approve specific servers from a project's `.mcp.json`

Turn individual plugins on or off per scope

Keep the `/model` Default choice inside your `availableModels` allowlist

Set environment variables for every session and its subprocesses

Show Claude's last response as comments when you press Ctrl+G to edit

Register marketplaces for a repository or an organization

Whether to automatically update this marketplace on Claude Code startup. Written automatically by Claude Code when you toggle auto-update for a marketplace

Local cache path where marketplace manifest is stored (auto-generated if not provided)

ISO 8601 timestamp of the last marketplace refresh. Written automatically by Claude Code

Where to fetch the marketplace from

Name backup models for when the primary is overloaded

Turn fast mode on for sessions where it's available

Require people to turn fast mode on each session

Control whether Claude queues feedback drafts for you to review

Change how often the session quality survey appears

Turn off or on the file snapshots that `/rewind` restores

Supply `@` file autocomplete from your own command

Make issue or review IDs in output into clickable links below the input box

Set the gateway URL the login screen connects to

Restrict login to claude.ai, Claude Console, or a cloud gateway

Pin claude.ai logins to your organization; only a managed source enforces it

Block startup until server-managed settings are freshly fetched

Refresh Google Cloud credentials with your own command

Run your own commands as hooks at points in Claude Code's lifecycle

Limit which env vars HTTP hooks can put in headers

Deprecated; use `attribution` to hide or change commit and PR attribution

Remove the built-in commit and PR instructions from the system prompt

Get a push notification when Claude is waiting on you

Ask you before Claude messages one of your sessions on another machine

Make `Ctrl+W` delete back to the previous whitespace, as Bash does

Have Claude respond in a language other than English

(Managed settings, third-party Desktop deployments only) MCP server configurations pushed to all users. Each entry specifies the transport and connection details, plus an optional toolPolicy map. Delivered through the managed settings file or MDM. See https://code.claude.com/docs/en/desktop#enterprise-configuration

Compose every managed source you deploy instead of using the highest-priority one alone

Keep auto-updates from installing anything below a version

Change the model Claude Code starts with

Map model IDs to your provider's IDs, such as Bedrock ARNs

Choose which models the `/model` picker lists, in your own order and with your own labels

Report spend at your organization's contracted rates instead of list price

Keep a saved effort level per model, which Claude Code writes when you run `/effort`

Generate rotating OpenTelemetry headers with your own command

Change Claude's role, tone, and output format with an output style

Apply or drop restrictions an SDK or IDE host passes when you deploy managed settings

Removed in v2.1.257, together with the `Ctrl+E` command explanation on shell permission prompts

Set allow, ask, and deny rules and the starting permission mode

Give Claude file access to directories outside the current one

Approve listed tool uses without a prompt

Always prompt before listed tool uses

Make the file tools refuse reads outside the working directories in every permission mode

Set the permission mode new sessions start in

Block listed tool uses, including reads of files that hold secrets

Prevent anyone from entering bypassPermissions mode

Choose where plan mode writes plan files

Store the answers you gave a plugin's configuration dialog

User configuration values for MCP servers keyed by server name

Non-sensitive option values from the plugin manifest's userConfig, keyed by option name. Sensitive values go to secure storage instead. See https://code.claude.com/docs/en/plugins-reference

Choose which marketplaces can surface plugin install suggestions in `/plugin`

Add your own text to the plugin trust warning

Run an executable that computes managed settings at startup

Name the helper executable Claude Code runs

Re-run the helper in the background on an interval

Set how long Claude Code waits for the helper

Point PR links at an internal code-review tool instead of github.com

Choose a terminal bell or desktop notification for task completion

Reduce or turn off spinner, shimmer, and flash animations

Run Claude Code's background processes through a corporate launcher on macOS and Linux

Choose the prompt cache lifetime for the main conversation

Hide the grayed-out prompt suggestions in the input box

Pick the default cloud environment for `claude --cloud`; a self-hosted `ccpool_` ID is read only from user and managed settings and `--settings`

Connect Remote Control automatically when a session starts

(Managed settings only) Run Claude's tools inside an isolated virtual machine instead of on your Mac. Disables the desktop app's iOS Simulator pane and Claude's simulator tools entirely, so the pane cannot attach a device while it is set. See https://code.claude.com/docs/en/desktop-ios-simulator#turn-off-simulator-access

Refuse to start on a version newer than your organization allows

Refuse to start on a version older than your organization requires

Keep gitignored files out of the `@` file picker

Stop Claude from responding after a `!` shell command runs

Isolate Bash commands from your filesystem and network on macOS, Linux, and WSL2

Let sandboxed commands send Apple Events on macOS

Let Claude retry a blocked command outside the sandbox, or forbid it

Run sandboxed commands without a permission prompt

Point the sandbox at a bubblewrap binary outside `PATH`

Hide or mask credential files and variables inside the sandbox

Let masked credentials reach plain HTTP services on trusted test networks

Link custom-named AWS key variables into one credential for re-signing

Unset or mask an environment variable inside the sandbox

Block or mask reads of a credential file inside the sandbox

Choose whether streaming, presigned, or SigV4A AWS requests fail or pass through

Run the Linux sandbox inside an unprivileged container

Let `gh`, `gcloud`, and `terraform` verify TLS behind a MITM proxy inside the sandbox on macOS

Turn on Bash sandboxing on macOS, Linux, and WSL2

Name commands that always run outside the sandbox

Refuse to start when the sandbox can't, instead of running unsandboxed

Control which paths sandboxed commands can read and write

Stop developers from re-opening read paths your organization blocked

Re-open reading inside a region `denyRead` blocks

Add paths sandboxed commands can write to

Block sandboxed commands from reading specific paths

Block sandboxed commands from writing to specific paths

Turn off filesystem isolation while keeping network isolation

Silence violation reports for paths a command is expected to probe

Control which hosts, ports, and sockets sandboxed commands reach

Let sandboxed commands connect to every Unix socket

Let sandboxed commands bind to localhost ports on macOS

Let macOS sandboxed tools like the iOS Simulator or Playwright reach their XPC services

Lock the network allowlist to managed settings

List Unix socket paths sandboxed commands can use on macOS

Pre-allow domains so sandboxed commands don't prompt for them

Block domains for sandboxed commands, even inside an allowed wildcard

Route sandbox HTTP traffic through your own proxy

Route sandbox SOCKS traffic through your own proxy

Deny hosts outside the allowlist instead of prompting

Have the sandbox proxy terminate TLS so it can read HTTPS requests

Use your own ripgrep binary inside the sandbox

Point the sandbox proxy at a `socat` binary outside `PATH`

Show a "clear context" option on the plan accept screen

See summaries of Claude's thinking instead of a collapsed stub

Hide the "Cooked for" duration after each response

Reserve more or less context for the skill listing

Cap each skill's description length in the skill listing

Hide or collapse a skill without editing its SKILL.md

Skip the one-time notice Claude Code shows when you first enter auto mode yourself rather than through the built-in default

Skip the confirmation dialog before bypassPermissions mode

Skip the WebFetch hostname check when Anthropic is unreachable

List of marketplace names the user has chosen not to install when prompted. See https://code.claude.com/docs/en/discover-plugins

List of plugin IDs (plugin@marketplace format) the user has chosen not to install when prompted. See https://code.claude.com/docs/en/discover-plugins

Underline misspelled words in the prompt input with a spell checker you install

Hide tips in the spinner while Claude works

Add your own tips to the spinner rotation, or replace the built-in tips

Add or replace the verbs shown while a turn runs

Add SSH connections to the Desktop environment dropdown

Limit which hosts Desktop SSH sessions can reach

Run your own command to render a status line below the prompt

Allowlist the marketplace sources users can add and install from

Block skills, agents, hooks, and MCP servers from user and project sources

Lock agents to plugin and managed sources

Lock hooks to plugin and managed sources

Lock MCP servers to plugin and managed sources

Lock skills to plugin and managed sources

Choose the prompt cache lifetime for subagents and other requests outside the main conversation

Rewrite rows in the subagent task display with your own command

Switch models automatically or pause when a safety classifier flags a request

Stop downloading the skills enabled on your claude.ai account and hide the ones already synced

Turn off syntax highlighting in diffs and code blocks

Removed in v2.1.234; see Specify teammates and models for how Claude Code picks a teammate's model

Choose how agent team teammates display

Hide the terminal progress bar in terminals that support it

Stop `/rename` and `--name` from changing the terminal tab title

Pick the interface color theme, built-in or custom

Show the times in the interface on a 12-hour or 24-hour clock, in UTC, or with a strftime pattern

Show the times in the interface in a time zone other than your system's

Choose the fullscreen or classic terminal renderer

Have Claude plan a workflow for each substantive task without being asked

Let the auto mode classifier review shell commands in plan mode; set `false` to get prompts instead

Show full tool output instead of truncated summaries; `viewMode` takes precedence when both are set

Start every session in default, verbose, or focus view

Map a two-key INSERT-mode sequence such as `jj` to Escape

Turn on voice dictation and pick hold or tap mode

Turn on voice dictation with the older single-key form

Turn off mouse-wheel acceleration in fullscreen rendering

Let the word `ultracode` in a prompt start a workflow; set `false` to type it without starting one

Set the agent count Claude aims for in dynamic workflows

Configure how Claude Code creates git worktrees

Branch new worktrees from the remote default branch or your local HEAD

Let background sessions edit the working copy without a worktree

Check out only the directories you need in each worktree

Symlink large directories into each worktree instead of duplicating them

Have WSL read managed settings from the Windows policy chain