CODE ECHO Burns my tokens unneeded. Abandoning Claude switching to / testing different brand.

Preflight Checklist

• [x] I have searched existing issues for similar behavior reports
• [x] This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Claude ignored my instructions or configuration

What You Asked Claude to Do

Problem

Claude is burning more tokens on compacting unwanted code listings then on new functionality.
The biggest problem:

IT FEELS VERY INTENTIONAL, To burn the money out of my pocket and make me upgrade.
I tried everything even with Claude-chat itself searching and advising me on the matter > Only real option: Spent more money on wasting energy and unused processing.

When using Claude Code on API billing, the model echoes back full code blocks after every file edit — even when instructed not to in CLAUDE.md with strong, capitalized, repeated instructions. This behavior persists across Opus and Sonnet models.
For multi-module projects (10+ files), this unnecessary output causes auto-compaction within a single enhancement task, burning tokens on content the developer never reads (code is reviewed in the editor — Zed, VS Code, etc.).
CLAUDE.md rules like "Do NOT output full code blocks" are acknowledged by the model and then immediately violated.
Impact

Doubles effective output token cost per interaction
Triggers premature auto-compaction, degrading session quality
Forces developers toward subscription upgrades for work that should be affordable on API billing
No workaround exists — this is not configurable in settings.json

Requested Solution
A settings.json property such as:
json{
"suppressCodeOutput": true
}
That instructs the model to respond with brief descriptions of changes only — not echoed code. The edit tool already writes to disk. The developer's editor already shows the diff. The terminal echo serves no purpose except burning tokens.
Environment

Claude Code CLI in Zed terminal
API billing (not subscription)
Opus and Sonnet models
Multi-module Python project

What Claude Actually Did

Huge listings of new and removed code.
Which where partly there unnecessary already == double token burning on no progress.

Expected Behavior

NO-Code echo at all.
Just synopsis.

Files Affected

Permission Mode

Accept Edits was ON (auto-accepting changes)

Can You Reproduce This?

Yes, every time with the same prompt

Steps to Reproduce

Easy, just use Claude.

Claude Model

Opus

Relevant Conversation

Impact

High - Significant unwanted changes

Claude Code Version

2.1.50 (Claude Code)

Platform

Anthropic API

Additional Context

I will now stop using Claude although the technical results are very pleasing.

This simple request feels so intentional and so obvious to get money burned for nothing. The Intelligents of Claude / opus is at a level that it absolutely must be intentional.