Free Claude Code settings precedence checker

Claude Code reads settings from up to four files plus the command line, and the same key can be set in several of them. This checker shows the value Claude Code ends up using for each key, which file it came from, which values were overridden, and which entries are ignored because they're in a file that can't set them. Paste only the files you have; leave the rest empty.

Everything runs in your browser. Nothing you paste is sent anywhere. Remove secrets (API keys, tokens in env) before you paste anyway.

Load an example with conflicts Clear everything

Problems and notes

Effective settings

Highest wins for single values: managed, then --settings, then local, then project, then user. Lists such as permissions.allow combine instead. Struck-through values were overridden.

Merged permission rules

Deny is checked first, then ask, then allow. The first match decides, and a narrower allow can't carve an exception out of a broader deny or ask.

Effective settings as one JSON object

Copy

This is a reading aid, not a file to save. Keep each setting in the file whose scope you want.

The rules this checker applies

• Order, highest first: managed settings, --settings JSON, .claude/settings.local.json, .claude/settings.json, ~/.claude/settings.json. Nothing in your own files or --settings overrides a managed key.
• Lists merge. permissions.allow, ask, deny, additionalDirectories and other arrays combine across files. Exceptions: fallbackModel is taken whole from the highest file; modelPicker is taken whole from managed, --settings or user settings and ignored in project and local files; a managed availableModels list is used as-is.
• env merges per variable, highest file wins for each variable. Project and local files can't set some variables (config and temp directories, telemetry exporters, and a few others).
• Hooks merge across files rather than replacing each other. enabledPlugins resolves per plugin. Same-name extraKnownMarketplaces entries are taken whole from the highest file.
• Some keys only work in some files. The checker knows the scope of every key in the official settings index: managed-only keys, user-or-managed keys, and global config keys that belong in ~/.claude.json. defaultMode values auto and bypassPermissions don't take effect from project or local settings.
• A few security keys keep their strict value even against managed settings: for example disableClaudeAiConnectors: true from any file, a project or local remoteControlAtStartup: false, and the lowest maxEffortLevel cap.
• Flags and shell variables are decided key by key, not by level: --model beats ANTHROPIC_MODEL, which beats the model key; CLAUDE_CODE_EFFORT_LEVEL beats --effort, which beats effortLevel; --permission-mode beats defaultMode unless bypass or auto mode is disabled.

What it can't know

• Objects the docs don't cover. For an object key like statusLine set in two files, the docs don't say whether fields merge or the whole object is replaced. The checker shows the highest file's object and says so.
• Your machine's state: whether you've trusted the folder (project allow rules wait for trust), whether settings.local.json is tracked by git, which managed source your organization uses, and your Claude Code version. Several rules above depend on the version.
• Every flag and variable. Only the overrides listed above are modeled. Each key's entry in the settings reference lists its own per-session overrides.

To confirm on your machine, run /status inside Claude Code (the Setting sources line lists which files loaded) and claude doctor in your terminal (it lists entries Claude Code rejected).

Setting still not applying? Read Claude Code settings not applying? 12 causes and fixes. Writing permission rules? Use the permissions builder, which tests a command against your rules. Hooks not firing? Build them with the hooks generator.

Rules on this page follow Anthropic's settings files and precedence, all settings and permissions docs. Not affiliated with Anthropic.

Made by Solo Stack Kits. If you'd rather start from finished examples, the free 3-skill sample has ready-made skills for client follow-ups and meeting notes, and the Solo Consultant Kit adds the full set.