Add compact display mode for MCP tool call outputs

Resolved 💬 3 comments Opened Mar 18, 2026 by dmacpherson Closed Mar 22, 2026

Summary

When using MCP tools (especially in multi-step workflows like document extraction), the Claude Code terminal becomes very noisy because it renders the full JSON input/output for every tool call. This makes it hard to follow the high-level progress of what's happening.

Current behavior

Every MCP tool call displays the full request parameters and full response body in the terminal. For tools that return large payloads (e.g., extraction instructions, chunk data, template lists), this can be hundreds of lines of output per call.

Desired behavior

A compact/summary display mode where MCP tool calls show:

  • Tool name + key parameters (e.g., add_document file.pdf, get_extraction_progress source_id=abc123)
  • A collapsed or hidden response body (expandable on demand)

The full data should still be passed to Claude — this is purely a UI/display concern.

Suggested approaches

  • A global setting like mcp_output_display: "compact" in settings.json
  • A per-tool or per-server config to control verbosity
  • Auto-collapse responses above a certain token threshold
  • A keybinding to toggle between compact and full output

Use case

Multi-step MCP workflows (e.g., document processing pipelines) that involve 10+ tool calls per run. The user wants to see progress at a glance ("indexed → planned → extracting → finalizing") without scrolling through walls of JSON.

🤖 Generated with Claude Code

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗