[FEATURE] Add "Explain" option to permission prompts in Claude Desktop code mode

Resolved 💬 2 comments Opened Mar 14, 2026 by aronwilson Closed Mar 17, 2026

Preflight Checklist

  • [x] I have searched existing requests and this feature hasn't been requested yet
  • [x] This is a single feature request (not multiple features)

Problem Statement

In Claude Code CLI, pressing Ctrl+E on a permission prompt lets you understand what a tool call will do before approving it — without breaking your flow. This is one of the CLI's best UX features for building trust and maintaining momentum.

In Claude Desktop's code mode, there's no equivalent. When you see a permission prompt and want to understand what it will do before approving, you have to:

  1. Deny the action
  2. Ask Claude to explain what it was trying to do
  3. Re-trigger the original action

This 3-step round trip kills momentum and is especially painful during longer agentic workflows where you're reviewing many tool calls.

  • CLI/Desktop parity: Desktop code mode should match CLI's permission UX, especially for power users who switch between both
  • Trust & safety: Understanding what you're approving is core to the human-in-the-loop model. Making it frictionless means people will actually use it instead of rubber-stamping Allow
  • Flow preservation: The current deny → ask → re-trigger loop is the single biggest UX gap between CLI and Desktop for agentic coding

Proposed Solution

Add an "Explain" option to the permission approval dialog in Claude Desktop code mode. This could be:

  • A third button alongside Allow / Deny (e.g., [Allow] [Deny] [Explain])
  • A ? icon or tooltip on the approval dialog
  • A keyboard shortcut (matching Ctrl+E from the CLI for muscle-memory parity)

When triggered, Claude would explain what the pending tool call does and why, then re-present the same approval prompt — no need to deny and re-trigger.

Alternative Solutions

_No response_

Priority

Medium - Would be very helpful

Feature Category

CLI commands and flags

Use Case Example

_No response_

Additional Context

_No response_

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗