Skip to content
AgentAtlas

Allow all commands in any AI coding agent: Cursor, Claude Code, Codex, Gemini CLI, Copilot CLI and 7 more (2026)

How do I let an AI coding agent run commands without asking for approval?

Every major coding agent has two levers: a full auto-approve switch (Run Everything in Cursor, --dangerously-skip-permissions in Claude Code, --yolo in Codex, Gemini CLI and Copilot CLI, YOLO Mode in Cline) and a scoped allowlist that pre-approves only the commands you trust. The allowlist is the safer default. Keep the full switch for containers, VMs or throwaway projects.

Also in: 简体中文 · 日本語 · 한국어 · Español · Português (Brasil)

Quick answer: the full auto-approve switch per tool

  • Cursor: Settings > Agents > Approvals & Execution, Run Mode = Run Everything (no sandbox, no review).
  • Claude Code: claude --dangerously-skip-permissions, meant only for containers or VMs. Auto mode is the reviewed middle ground: no routine prompts while a classifier checks actions.
  • OpenAI Codex CLI: --dangerously-bypass-approvals-and-sandbox, alias --yolo, removes both approvals and the sandbox.
  • Gemini CLI: --yolo (same as --approval-mode yolo). It can only be turned on from the command line.
  • GitHub Copilot CLI: --allow-all-tools skips tool prompts; --allow-all or --yolo turns on all permissions at once.
  • Cline: Cline Settings > Features > YOLO Mode. It disables all safety checks.
  • goose: GOOSE_MODE=auto, which is already the default and edits and deletes files without asking.
  • Junie CLI: brave mode, toggled with /brave or Ctrl+B, or junie --brave. Level On runs all sensitive actions without asking.
  • Factory Droid: droid exec --auto high for headless runs.
  • OpenCode: set permission rules to allow in opencode.json.
  • Kiro: allow rules in ~/.kiro/settings/permissions.yaml.
  • Zed: agent.tool_permissions with default set to allow.

The safer way: pre-approve only trusted commands

Most tools let you allow specific commands while still asking for everything else. This keeps routine work like tests, linters and git reads fast without handing the agent rm, git push or cloud CLIs.

Claude Code: .claude/settings.json

{
  "permissions": {
    "allow": ["Bash(npm run *)", "Bash(git commit *)"],
    "deny": ["Bash(git push *)"]
  }
}

Cursor: .cursor/permissions.json

{
  "terminalAllowlist": ["git", "npm:install*", "cargo"],
  "mcpAllowlist": ["github:*", "linear:list_issues"]
}

Gemini CLI: .gemini/settings.json

{
  "tools": {
    "allowed": ["run_shell_command(git)", "run_shell_command(npm test)"]
  }
}

GitHub Copilot CLI: allow everything except rm and git push

copilot --allow-all-tools --deny-tool='shell(rm)' --deny-tool='shell(git push)'

OpenCode: opencode.json (last matching rule wins)

{
  "permission": {
    "bash": {
      "*": "ask",
      "git *": "allow",
      "npm *": "allow",
      "rm *": "deny"
    }
  }
}

Kiro: ~/.kiro/settings/permissions.yaml

rules:
  - capability: shell
    match: ["git *", "npm *", "npx *"]
    effect: allow

Sandboxes: auto-approve with a safety net

  • Codex pairs an approval policy with a sandbox: read-only, workspace-write or danger-full-access. codex --sandbox workspace-write --ask-for-approval on-request lets it edit the workspace and asks before leaving it. Network access is off by default in workspace-write.
  • Cursor's Auto-review mode runs allowlisted calls immediately and sends other shell commands to the sandbox when possible.
  • Junie's --sandbox flag lets you use brave mode with less risk; it is only in development, nightly and experimental builds and not on Windows.
  • Claude Code's own guidance is to use --dangerously-skip-permissions only inside a container or VM, as a non-root user.

Which rule wins when allow and deny overlap

  • Claude Code: deny first, then ask, then allow.
  • GitHub Copilot CLI: --deny-tool always beats the allow options.
  • Kiro: deny beats ask, and ask beats allow, across every scope.
  • OpenCode: the last matching rule wins, so put the catch-all first and specific patterns after it.
  • goose: environment variables override config.yaml, which overrides defaults.

Step-by-step guides per tool

Sources

More Claude Code guides

Reviewed Oct 23, 2026. Settings change often; the linked vendor docs are the source of truth.

Built an MCP server? Get it listed everywhere →

Shipping an Android app? Google Play needs 12 testers for 14 days first. Get them with Tester Exchange →