Settings Overview

Claude Code settings.json Reference

Per-section reference for every recognized key in settings.json.

Compiled from: plugin refs (claude-code-internals v2.5.0), the CC binary bundle (v2.1.105), local settings.json files on this machine, and the official docs at code.claude.com/docs/en/settings. ~125+ keys total.

Table of Contents

1. Attribution & Output
2. Authentication & API
3. Channel & Communication
4. Configuration & Environment
5. Development Tools & IDE
6. Enterprise & Organization
7. File & Directory Handling
8. File System Sandbox
9. Hooks & Automation
10. LLM & Model
11. Memory & Context
12. MCP Servers
13. Network & Proxy
14. Permissions & Security
15. Plugins & Extensions
16. Subagents
17. UI & Display
18. Worktree
19. Telemetry / Misc

Overview

19 sections, 119 keys total.

#	Section	Description	Keys
1	Attribution & Output	Git commit and PR attribution trailers (e.g., Co-Authored-By lines).	3
2	Authentication & API	API-key helpers, AWS credential refresh, forced login methods, and OTEL auth helpers.	7
3	Channel & Communication	Voice I/O and the managed channel-plugin allowlist.	3
4	Configuration & Environment	Injected env vars, default shell, and UI language.	3
5	Development Tools & IDE	IDE auto-connect, extension install, editor mode, gitignore respect, and git instruction injection.	5
6	Enterprise & Organization	Managed-only lockdowns for hooks, MCP servers, permission rules, plus announcements and marketplace blocklists.	6
7	File & Directory Handling	Auto-memory and /plan output directories, plus custom @-mention file-suggestion commands.	4
8	File System Sandbox	Read/write allow/deny paths, process sandbox toggles, and weaker-sandbox escape hatches for Docker/macOS.	13
9	Hooks & Automation	PreToolUse/PostToolUse/SessionStart/Stop hook definitions and HTTP-hook URL/env allowlists.	5
10	LLM & Model	Default model, enterprise model allowlist, per-tool overrides, thinking/effort tuning, and mode hard-disables.	9
11	Memory & Context	Auto-memory, background consolidation, transcript retention, compaction thresholds, and edit checkpointing.	5
12	MCP Servers	MCP server config, allow/deny lists, and .mcp.json enable/disable toggles.	6
13	Network & Proxy	Outbound domain allowlist, HTTP/SOCKS proxy ports, Unix socket access, and macOS Mach/XPC services.	8
14	Permissions & Security	Tool-use allow/deny/ask rules, default permission mode, auto-mode classifier, and danger-mode prompt skips.	13
15	Plugins & Extensions	Enabled plugins, extra marketplaces, plugin trust messaging, and plugin-only customization lock.	4
16	Subagents	Custom agent definitions and main-thread-as-subagent execution.	2
17	UI & Display	Theme, status line, output/view mode, thinking summaries, spinner tips, progress bars, and update channel.	16
18	Worktree	Worktree symlink directories and Git sparse-checkout paths.	2
19	Telemetry / Misc	Feedback survey rate, teammate mode, deep-link registration, resource-unavailable behavior, and claude.ai linkage.	5

Quick reference — all settings keys

Key	Category	Type	Description
advisorModel	LLM & Model	string	Secondary advisor model (v2.1.90)
agent	Subagents	string	Run main thread as named subagent
agents	Subagents	object	Custom agent definitions
allow / deny / ask / defaultMode	Permissions & Security	—	Backwards-compat shorthands for the above
allowedChannelPlugins	Channel & Communication	array	Managed-only. Allowlist of channel plugins
allowedHttpHookUrls	Hooks & Automation	array	URL allowlist for HTTP hooks (wildcards ok)
allowedMcpServers	MCP Servers	array	Allowlist (name, command, glob URL)
allowManagedHooksOnly	Enterprise & Organization	bool	Managed-only. Block user/project/plugin hooks
allowManagedMcpServersOnly	Enterprise & Organization	bool	Managed-only. Only managed MCP servers
allowManagedPermissionRulesOnly	Enterprise & Organization	bool	Managed-only. Ignore user/project permission rules
allowUnsandboxedCommands	File System Sandbox	array	Commands allowed to run outside sandbox
alwaysThinkingEnabled	LLM & Model	bool	Extended thinking on by default
apiKeyHelper	Authentication & API	object	API key storage/retrieval mechanism
attribution.commit	Attribution & Output	string	Git commit trailer (e.g., “Co-Authored-By: Claude…”). Empty string disables
attribution.pr	Attribution & Output	string	PR description attribution. Empty string disables
autoCompactWindow	Memory & Context	number	Token threshold for compaction
autoConnectIde	Development Tools & IDE	bool	Auto-connect to IDE if available
autoDreamEnabled	Memory & Context	bool	Background consolidation
autoInstallIdeExtension	Development Tools & IDE	bool	Auto-install IDE extensions
autoMemoryDirectory	File & Directory Handling	string	Auto-memory dir path
autoMemoryEnabled	Memory & Context	bool	Toggle auto-memory extraction
autoMode	Permissions & Security	object	Auto-mode classifier configuration
autoUpdatesChannel	UI & Display	string	stable / beta / …
availableModels	LLM & Model	array	Enterprise allowlist
awsAuthRefresh	Authentication & API	bool	Auto-refresh AWS session credentials
awsCredentialExport	Authentication & API	string	AWS credential export config
blockedMarketplaces	Enterprise & Organization	array	Marketplace blocklist
channelsEnabled	Channel & Communication	bool	Enable channel communication features
claudeai	Telemetry / Misc	string	claude.ai connection identifier
cleanupPeriodDays	Memory & Context	number	Transcript retention window
companyAnnouncements	Enterprise & Organization	array	Company announcements to display
defaultShell	Configuration & Environment	string	Default shell (bash/zsh/fish/…)
deniedMcpServers	MCP Servers	array	Blocklist
disableAllHooks	Hooks & Automation	bool	Managed-only. Disable all hooks including managed
disableAutoMode	LLM & Model	bool	Hard-disable auto mode
disableBypassPermissionsMode	LLM & Model	bool	Hard-disable bypass-permissions mode
disableDeepLinkRegistration	Telemetry / Misc	bool	Disable deep-link protocol registration
disabledMcpjsonServers	MCP Servers	array	Disabled servers from .mcp.json
disableSkillShellExecution	Hooks & Automation	bool	Disable shell execution inside skills
editorMode	Development Tools & IDE	string	Editor UI mode (“fullscreen”, “pane”, …)
effortLevel	LLM & Model	string	fast / balanced / thorough (also low/medium/high/max)
enableAllProjectMcpServers	MCP Servers	bool	Auto-enable all project MCP servers
enabledMcpjsonServers	MCP Servers	array	Enabled servers from .mcp.json
enabledPlugins	Plugins & Extensions	object	Marketplace-scoped plugin ID → bool
enableWeakerNestedSandbox	File System Sandbox	bool	Weaker sandbox for unprivileged Docker (Linux/WSL2)
enableWeakerNetworkIsolation	File System Sandbox	bool	Allow system TLS trust service (macOS; ⚠ weakens security)
env	Configuration & Environment	object	Env vars injected into all sessions
extraKnownMarketplaces	Plugins & Extensions	object	Additional marketplaces beyond built-ins
failIfUnavailable	Telemetry / Misc	bool	Fail if requested resource unavailable
fastModePerSessionOptIn	LLM & Model	bool	Allow per-session fast-mode opt-in
feedbackSurveyRate	Telemetry / Misc	number	Survey sampling rate
fileCheckpointingEnabled	Memory & Context	bool	Snapshot files before edits (default true)
fileSuggestion.command	File & Directory Handling	string	Custom command for @-mention autocomplete
fileSuggestion.type	File & Directory Handling	string	”command” or built-in
filesystem.allowManagedReadPathsOnly	File System Sandbox	bool	Managed-only.
filesystem.allowRead	File System Sandbox	array	Read-allowed paths
filesystem.allowWrite	File System Sandbox	array	Write-allowed paths
filesystem.denyRead	File System Sandbox	array	Read-denied paths
filesystem.denyWrite	File System Sandbox	array	Write-denied paths
forceLoginMethod	Authentication & API	string	Force login method (github, google, …)
forceLoginOrgUUID	Authentication & API	string	Force org UUID for login
forceRemoteSettingsRefresh	Authentication & API	bool	Force remote managed settings refresh
hooks	Hooks & Automation	object	PreToolUse / PostToolUse / SessionStart / Stop / CwdChanged / Notification
httpHookAllowedEnvVars	Hooks & Automation	array	Env vars HTTP hooks may use in headers
includeCoAuthoredBy	Attribution & Output	bool	Deprecated — use attribution.commit
includeGitInstructions	Development Tools & IDE	bool	Include git instructions in system prompt
language	Configuration & Environment	string	UI language
mcpServers	MCP Servers	object	MCP server configurations
model	LLM & Model	string	Default model (e.g., opus[1m], sonnet-4-6)
modelOverrides	LLM & Model	object	Per-skill / per-tool model overrides
network.allowAllUnixSockets	Network & Proxy	bool	Allow all Unix socket access
network.allowedDomains	Network & Proxy	array	Outbound domain allowlist (wildcards)
network.allowLocalBinding	Network & Proxy	bool	Allow binding to local ports
network.allowMachLookup	Network & Proxy	array	Allowed XPC/Mach services (macOS)
network.allowManagedDomainsOnly	Network & Proxy	bool	Managed-only.
network.allowUnixSockets	Network & Proxy	array	Allowed Unix socket paths
network.httpProxyPort	Network & Proxy	number	Custom HTTP proxy port
network.socksProxyPort	Network & Proxy	number	Custom SOCKS5 proxy port
otelHeadersHelper	Authentication & API	object	OpenTelemetry headers helper
outputStyle	UI & Display	string	Output rendering style
permissions.allow	Permissions & Security	array	Allow rules (e.g., Bash(npm run *), Read(*.ts))
permissions.ask	Permissions & Security	array	Confirm-before-running rules
permissions.defaultMode	Permissions & Security	string	auto / bypassPermissions / ask / plan
permissions.deny	Permissions & Security	array	Deny rules
plansDirectory	File & Directory Handling	string	Where /plan outputs are saved
pluginTrustMessage	Plugins & Extensions	string	Custom trust prompt message
prefersReducedMotion	UI & Display	bool	Respect reduced-motion preferences
respectGitignore	Development Tools & IDE	bool	Respect .gitignore in file suggestions
sandbox.autoAllowBashIfSandboxed	File System Sandbox	bool	Auto-allow Bash in sandbox mode
sandbox.enabled	File System Sandbox	bool	Enable process sandbox
sandbox.excludedCommands	File System Sandbox	array	Commands excluded from sandbox
sandbox.filesystem	File System Sandbox	object	Nested FS sandbox rules
sandbox.network	File System Sandbox	object	Nested network sandbox rules
showClearContextOnPlanAccept	UI & Display	bool	Show clear-context after plan accept
showThinkingSummaries	UI & Display	bool	Show extended-thinking summaries
showTurnDuration	UI & Display	bool	Show per-turn duration
skipAutoPermissionPrompt	Permissions & Security	bool	Skip auto-mode opt-in prompt
skipDangerousModePermissionPrompt	Permissions & Security	bool	Skip “danger mode” confirmation dialog
spinnerTipsEnabled	UI & Display	bool	Spinner tips during ops
spinnerTipsOverride	UI & Display	string	Override default spinner tips
spinnerVerbs	UI & Display	array	Custom spinner verbs
statusLine.command	UI & Display	string	Custom status line command
statusLine.type	UI & Display	string	"command" or built-in
strictKnownMarketplaces	Enterprise & Organization	bool	Managed-only. Restrict plugin marketplaces to known list
strictPluginOnlyCustomization	Plugins & Extensions	bool	Lock skills/agents to plugin delivery
teammateMode	Telemetry / Misc	bool	Teammate collaboration mode
terminalProgressBarEnabled	UI & Display	bool	Show terminal progress bars
theme	UI & Display	string	UI theme preference
useAutoModeDuringPlan	Permissions & Security	bool	Enable auto-mode during /plan execution
viewMode	UI & Display	string	View mode preference
voiceEnabled	Channel & Communication	bool	Enable voice I/O
worktree.sparsePaths	Worktree	array	Git sparse-checkout paths for worktrees
worktree.symlinkDirectories	Worktree	array	Directories to symlink into worktrees

See Also

• ../ENV/README.md — companion reference for environment variables.
• ../Commands/README.md — companion reference for built-in slash commands.
• ../Skills/README.md — companion reference for bundled skills.
• ../Tools/README.md — companion reference for built-in tools.
• Official docs: https://code.claude.com/docs/en/settings