[FEATURE] Middle display level: show tool command + first N output lines (between collapsed and verbose)

Status Open
Reported on v2.1.282
Maintainer reply None cached
Activity 0 comments · opened Sep 25, 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

Tool-call display in the interactive TUI is currently all-or-nothing:

  • verbose: false — consecutive tool calls are grouped into a single line like Ran 2 shell commands. I can no longer see which command was run without pressing Ctrl+O.
  • verbose: true — every command and its full output is printed inline, which floods the screen during long agentic sessions and buries the assistant's short explanations between tool calls.

What I actually want to follow along is the level in between: the command itself, plus the first few lines of its output. The command text is essential context (what is Claude doing to my system?), while the full output is usually noise unless something looks off.

Related but different requests:

  • #21569 — same complaint for the Read tool (Read (3 files) hides the file paths); closed as not planned.
  • #91827 / #40428 — ask for a persistent collapsed-by-default setting. That is still a binary choice; this request is about a middle level of detail.

The rendering pieces already exist — the … +N lines (ctrl+o to expand) preview is shown in some states — so this is mostly about exposing it as a selectable default.

Proposed Solution

Add a third display level (or a numeric preview setting), for example in settings.json:

{
  // "collapsed" | "preview" | "verbose"
  "toolOutputDisplay": "preview",
  // number of output lines shown per tool call in "preview" mode
  "toolOutputPreviewLines": 5
}

In preview mode:

  1. Each tool call is shown on its own line with its key argument (the full Bash command, the file path for Read/Edit, the MCP tool name + short params). No grouping into Ran N shell commands.
  2. Below it, the first N lines of output, followed by … +M lines (ctrl+o to expand).
  3. Ctrl+O keeps working as today to toggle the full transcript view.

Exposing it in /config next to "Verbose output" would make it easy to discover.

Alternative Solutions

  • Toggling Ctrl+O back and forth — works, but it switches the whole view and loses my place, and I have to do it constantly just to see the command names.
  • Keeping verbose: true — shows the commands, but long outputs make the conversation hard to read.
  • Asking Claude (via CLAUDE.md) to quote the commands and key output lines in its text — works only on a best-effort basis, costs tokens, and duplicates what the UI already knows.

Priority

Medium - Would be very helpful

Feature Category

Interactive mode (TUI)

Use Case Example

  1. I ask Claude to reconfigure some plugins / MCP servers on my machine; it runs ~10 shell commands (backups, claude mcp remove, package installs, verification).
  2. In non-verbose mode I only see Ran 1 shell command, Ran 2 shell commands, … and have to trust the prose summary or keep pressing Ctrl+O.
  3. With preview mode I would see e.g. Bash(sudo pacman -S pyright) plus 3–5 lines of output inline — enough to verify what's happening at a glance, without the wall of text of verbose mode.

---
Environment: Claude Code 2.1.282, Linux (Arch), kitty terminal.

View original on GitHub ↗