# How to stop Claude Code and Codex asking for permission every time (safely)

> Claude Code and Codex keep asking for approval. Three ways to cut the prompts, from pre-approving safe commands to turning approval off, and which are worth it.

Canonical: https://herdbell.com/blog/stop-ai-coding-agent-asking-permission
Author: Herdbell team
Published: 2026-10-27
Last updated: 2026-10-27

Claude Code and Codex ask before they run commands that can change things, and when you are doing real work that feels like constant interruption. There are three ways to cut it down. They are not equal: the first two keep you in control, and the last gives it up.

## Three levels, safest first

| Level | What you do | Risk |
|---|---|---|
| 1. Pre-approve the safe commands | Add allow rules in Claude Code's permission settings, or pick a lighter approval policy in Codex | Low: only commands you chose run unasked |
| 2. Answer faster | Keep prompts on and answer each in one keystroke from any window with Herdbell | Low: you still see every command |
| 3. Turn approval off | Claude Code's skip flag, or Codex's approval policy set to never | High: commands run without you seeing them |

Setting names and options vary by version, so check each agent's own documentation. Claude Code has permission modes and allow rules in its settings; Codex has an approval policy with options such as untrusted, on-request and never.

## What I would do

1. **Start with level 1.** Pre-approve the read-only and routine commands you are always asked about, such as listing files or running your tests.
2. **Add level 2 for the rest.** With Herdbell, the remaining requests show as a card at your cursor and you press **Return** to allow or **D** to deny. Risky commands ask twice.
3. **Keep level 3 for throwaway environments only,** such as a disposable container or a project you can restore from git in seconds. See [why skipping is risky](https://herdbell.com/blog/claude-code-dangerously-skip-permissions-alternative).

## What Herdbell does and does not do

Herdbell never turns approval off and never approves anything for you. It makes the answer take a second, and falls back to the agent's own prompt if you do not answer in time.

## Common questions

### How do I make Claude Code stop asking for permission?
Pre-approve the commands you trust in its permission settings, and answer the rest quickly. Turning approval off entirely is possible but means commands run without you seeing them.

### Is setting the Codex approval policy to never safe?
Only in an environment you can throw away or restore. With never, Codex runs commands without asking, so a wrong command is not stopped.

## Related
- [Herdbell: pricing](https://herdbell.com/#pricing)
- [All posts](https://herdbell.com/blog)
- [Skipping Claude Code permissions is risky: a safer way to move faster](https://herdbell.com/blog/claude-code-dangerously-skip-permissions-alternative)
- [AI coding agent safety checklist: 10 rules before you let it run commands](https://herdbell.com/blog/ai-coding-agent-safety-checklist)