[FEATURE] Always-visible / always-expanded tool call output in VS Code extension

Resolved 💬 2 comments Opened May 26, 2026 by VirtualL Closed Jul 7, 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

When Claude Code runs shell commands (e.g. kubectl, git, helm, npm) inside the VS Code
extension, the command and its output appear as collapsed blocks in the chat panel.
To see what was run and what it returned, I must manually click each block to expand it.

When working with infrastructure tooling (Kubernetes, Terraform, etc.), this is a
safety problem — LLMs can run destructive or unexpected commands (deleting resources,
modifying cluster state), and I need to passively monitor what is being executed in
real time without extra clicks interrupting my workflow.

Proposed Solution

Add a VS Code setting (e.g. claude.alwaysExpandToolCalls: true) or a persistent toggle
in the extension UI that keeps tool call blocks expanded by default, so commands and
their full output are always visible without manual interaction.

Alternative Solutions

  • Using claude --verbose in the terminal works, but loses the VS Code chat UI experience.
  • Clicking each block manually is the current workaround, but it's disruptive and easy

to miss output that already scrolled by.

VS Code addressed the same gap for GitHub Copilot agent mode in
microsoft/vscode#257468, where inline terminal output with expand/collapse was implemented.

Priority

Critical - Blocking my work

Feature Category

Configuration and settings

Use Case Example

  1. I ask Claude Code to debug a Kubernetes deployment issue
  2. Claude runs several kubectl get / kubectl describe / kubectl logs commands
  3. With collapsed output, I have no visibility into what's being queried or returned

unless I click every block

  1. With always-expanded output, I can follow along in real time and catch if Claude

is querying the wrong namespace or misreading a resource before it acts on it

Additional Context

_No response_

View original on GitHub ↗

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