Claude Code Permissions and Settings for Solo Developers

Claude Code supports fine-grained permissions so that you can specify exactly what the agent is allowed to do and what it can't. Permission rules are enforced by Claude Code, not by the model. Instructions in your prompt or CLAUDE.md shape what Claude tries to do, but they don't change what Claude Code allows.

Manual Mode and Approval Workflows

In Manual mode, Claude Code stops and asks you before most actions that edit files, run shell commands, or reach the network.

Settings are the JSON keys that change how Claude Code behaves: which model it starts with, what it can run without asking, which files it can't read, how it looks in your terminal, and what your organization enforces. Claude Code reads settings from four files, and an organization can also deliver managed settings from the claude.ai console. Each source has a scope: the set of people and projects a setting saved in it applies to, whether that's only you, everyone in a project, or everyone in your organization.

Handling Conflicting Settings

When the same key appears in more than one place, Claude Code uses the value from the highest level that sets it. When you set the same list key, such as permissions.allow, in more than one file, Claude Code combines the lists instead of picking one, so each file can add entries without removing another file's.

Managing Permissions via CLI

You can view and manage Claude Code's tool permissions with /permissions. The dialog lists all permission rules and the settings.json file each rule comes from.

For sensitive work, a solo developer can stay in Manual mode, where Claude Code asks before most actions that edit files, run shell commands, or reach the network. Switch permission modes with Shift+Tab in the CLI. Prompt instructions can't change what Claude Code allows, so set boundaries in settings. A deny rule in your user settings (~/.claude/settings.json) applies in every project on this machine, and no other level can allow that tool. For keys that aren't lists, precedence runs from managed settings to command line, project local, shared project and then user settings, so if a value in your user settings doesn't take effect, check whether a higher level sets the same key. /permissions shows the settings.json file each rule comes from.

The observation window is 2026-10-05 through 2026-11-04. Last30Days discovered 10 items; 8 were retained after requiring a direct mention of the topic, an HTTPS URL, a substantive excerpt, and deduplication. Included/discovered counts by source are reddit: 0/1, x: 7/8, youtube: 1/1. Engagement was known for 1/8 retained items; no engagement was inferred for the remainder. This section reports public posts and videos observed in the window, not vendor policy or a market-wide statistic. The sample is not representative of the whole market, and each observation links to its source below.

• 2026-11-03 · x: Claude Code still asks before a few especially risky rm commands. you can now set CLAUDE_CODE_EVEN_MORE_DANGEROUSLY_SKIP_PERMISSIONS=1. (source)
• 2026-11-02 · youtube: How Claude Code permissions work: the four modes (Manual, Accept edits, Plan, Auto), allow rules that stop the prompts for commands you trust, and deny rules that block what must never run. The allow and deny rules from the video, in .claude/settings.json: { "permissions": { "allow": ["Bash(pnpm test *)", "Bash(pnpm run *)"], "deny": ["Bash(git push *)",… (source)
• 2026-10-30 · x: Claude Code 2.1.314 or later has a setting that makes Write, Edit and NotebookEdit refuse any path outside your working directories. (source)
• 2026-10-30 · x: New security setting in Claude Code 2.1.314 You can now prevent Claude from writing files outside your working directories, even in bypass permissions mode ... (source)
• 2026-10-29 · x: Older versions don't recognize it and skip the whole settings file, so the permissions and hooks you set up in it stop working too. Run claude --version first. (source)
• 2026-10-20 · x: Claude Code ALWAYS ask for your approval before executing these six sensitive Stripe tools, Settings for Claude Code , currency and recipient. (source)
• 2026-10-18 · x: Claude Code asking permission before every file edit is safer. Middle ground works for me: edits allowed, bash commands ask first. Permissions control what ... (source)
• 2026-10-15 · x: Armor makes Claude Code's auto permissions mode compatible with any model. For instant setup, run Claude Code and your favorite model with Ori: $ ori claude ... (source)