Answer Claude Code permission prompts from the notch.
AgentPeek surfaces Claude Code and Codex permission requests right where you work, so you can allow or deny without hunting for the right terminal.
Inline permission prompts, not buried terminals
Claude Code and Codex pause to ask before they run a tool, edit a file, or execute a command. When you have several sessions open, finding the one that is waiting wastes time and breaks flow. AgentPeek opens the prompt inline in the panel, right next to the session it belongs to, with the full detail of what the agent wants to do.
Keyboard shortcuts for allow and deny
Decide fast without reaching for the mouse:
Allow ⌘A
Approve the request and let the agent keep going.
Deny ⌘N
Block the action; add feedback so the agent can adjust.
Always Allow ⌥A
Permit this kind of action from now on (Claude Code only).
Open Terminal ⌥T
Jump straight to the session's terminal window.
A Deny with feedback button lets you tell the agent why, so it can adjust instead of just stopping.
Approve plans and reviews too
Permission prompts are not the only thing agents wait on. For plan and review prompts the buttons read Approve and Reject, so you can sign off on a plan before the agent starts executing. AgentPeek queues pending prompts and lets you move between them, so nothing waits silently in a background terminal.
You stay in control
AgentPeek only responds when you choose an action. Prefer to keep approvals in the terminal? The prompt-handling setting lets you leave them there and use AgentPeek purely for visibility. Either way, the decision is always yours.
Works across all your sessions
Permission prompts from any Claude Code or Codex session land in the same panel, alongside live session state and token usage. See the full reference in the docs.
Frequently asked questions
Approve agents without losing flow.
Free for 2 days, then a $15 one-time license. Activate once during onboarding.
Download for Mac