Blog
Guides

Can you route a Claude Code permission prompt to your team lead before it runs?

Claude Code asks the person at the terminal. Three ways to get a team lead's yes before a push or deploy runs instead, and what each one can't do.

Aadil Ghani
Aadil Ghani
Founder, Pushary
Nov 7, 20264 min read
Share

Not with Claude Code alone. A permission prompt goes to the person running the session, at their terminal or on their own phone. You can force Claude Code to ask before a git push, but you can't choose who answers. To get a lead's yes first, you need a hook that waits for someone else, or a tool that routes the prompt for you.

Everything below was checked against Claude Code's docs on 7 November 2026.

Key takeaways

  • Claude Code sends every permission prompt to the person at that terminal. No setting names another person.
  • Managed settings can make a step ask. They can't decide who is asked.
  • A PermissionRequest hook can hold the step and wait up to 600 seconds by default for someone else's answer, if you build that part.
  • A Pushary routing rule sends a matching step to a lead or a role, on their phone and in the dashboard, and records who said yes. It doesn't lock the terminal.

What Claude Code does on its own

A permission prompt asks whoever is at that terminal. Remote Control moves the prompt to the same person's phone. Channels, a research preview, relays it to Telegram, Discord or iMessage, but anyone who can reply there can approve, and the reply carries no name.

Managed settings are the team-wide rules. They can put Bash(git push:*) in the ask list so every push stops for a yes. The yes still comes from the person who ran the agent.

So on a team, the developer whose agent wants to push to the infrastructure repository is the one who approves it. If that's not who you want deciding, Claude Code has no setting for it. Who approved it? checks the same question for Codex, Cursor and Copilot. None of them has one either.

Build it with a hook

Claude Code runs a PermissionRequest hook when a tool call needs a decision. The hook gets the tool name and its input, and it can return a decision of allow or deny. A command hook can run for up to 600 seconds by default before Claude Code cancels it, and you can set a different timeout.

That's enough to build the basic version:

  1. Match the steps you care about, such as Bash calls that start with git push.
  2. Post the request somewhere the lead will see it: a Slack message with two buttons, or a small web page.
  3. Wait for the answer, then return allow or deny.

What you'd still have to build:

  • Knowing who answered. A Slack button press is easy to attribute. A link in a chat isn't.
  • What happens at the timeout. A hook that times out renders no decision, so Claude Code falls back to its own prompt, and the person at the terminal decides after all.
  • A record. Something has to keep which step was asked, who answered and when.
  • Every machine. The hook has to be on each teammate's computer, for each agent they run. Codex and Cursor have their own hook formats.

If your team runs only Claude Code and you're happy to maintain the script, this works.

Route it with Pushary for Teams

Pushary for Teams does the routing without a script. Each person connects their own agents. Once a second person joins, each question goes to the person whose agent asked, and an owner or admin can send a chosen step to someone else.

In the dashboard, open Policies, find Approval routing and add a rule:

  • Tool pattern: Bash(git push:*) for every git push.
  • Repository: infra, or all repositories.
  • Route to: a role such as Admins, or one person.

Now when Tom's Claude Code wants to push to infra, Pushary asks the admins on their phones and in the dashboard instead of Tom. Pushes to other repositories still ask Tom. Only they can answer is on by default, so Pushary refuses an answer from anyone else. Whoever answers in the Pushary phone app, Mac app or dashboard is saved by name, next to whose agent asked.

The same rule works for Codex, Cursor, Gemini CLI, OpenCode and Hermes, so a team that runs more than one agent keeps one set of rules and one record.

What a routing rule can't do

  • It doesn't lock the terminal. If the lead hasn't answered when the wait ends, the step goes back to Claude Code, and the person who started it can say yes at the prompt. That yes isn't recorded.
  • The phone can't approve a risky step. A plain push can be approved from the phone. A force push, rm -r or a deploy can only be denied there; approve those in the dashboard.
  • It matches tool approvals only. A question the agent writes itself, such as which of two migrations to keep, goes to the person whose agent asked.
  • It needs a second person. Routing starts once someone else joins the workspace.

The Team docs keep these limits current. Team is $59 a month for 5 people, plus $12 a month for each extra person, and new workspaces get 3 days free.

Frequently asked questions

Does Claude Code have a setting that sends a permission prompt to someone else?

No. As of 7 November 2026, a Claude Code permission prompt goes to the person running the session, at the terminal or on their own phone through Remote Control. Managed settings can force a prompt for a step such as git push, but they cannot choose who answers it. Channels relays prompts to Telegram, Discord or iMessage, but anyone in the channel can approve and the reply carries no name.

Can a PermissionRequest hook wait for another person's answer?

Yes, if you build the waiting part. A PermissionRequest hook runs when a tool call needs a decision and can return allow or deny. A command hook may run for up to 600 seconds by default, so a script can post the request to Slack or a web page, wait for a lead's answer and return it. You then own the identity check, the timeout, the record and keeping the hook on every teammate's machine.

Can the person at the terminal still say yes when a lead is supposed to decide?

With Pushary's routing rules, yes. A rule decides who Pushary asks, and Only they can answer refuses an answer from anyone else in Pushary, but it does not lock the terminal. If the lead has not answered when the wait ends, the step goes back to Claude Code, and the person who started it can approve it at the prompt. That yes is not recorded.

Aadil Ghani
Aadil Ghani
Founder, Pushary

Building Pushary so an AI agent can reach you on your phone and wait for a yes before it does something you would not want.

Read next

Guides

Tools that let a team approve an AI coding agent's risky steps, compared

Claude Code, Codex and Cursor only ask the person at the keyboard. These tools can ask a lead instead. Which agents each one covers, where approvers answer, what it records and what five people pay.

Nov 7, 20266 min readAadil Ghani
Guides

Claude Code managed settings for a team: a working file, and what it can't do

The one Claude Code file your team can't override: where it goes, a version you can copy, how its rules are read, and what it still can't do for a team.

Nov 6, 2026Updated Nov 7, 20266 min readAadil Ghani
Guides

An AI agent broke something. How to trace what ran and whose yes let it

Git names the machine, the session names the command, and only a recorded approval names who said yes. Where to look, in order, and what each place can and can't tell you.

Nov 7, 20263 min readAadil Ghani

Let your agents work. Stay in control.

Get the questions that need you on your phone or Mac. Approve, reject, or answer so your agent knows what to do next.