claude-code - 💡(How to fix) Fix Feature: Show token usage after every prompt [1 comments, 2 participants]

Summary

Add an option to display token usage statistics (input tokens, output tokens, cache hits, cost) automatically after each response, without requiring the user to manually open the /usage dialog.

Motivation

Currently, the only way to view token usage is via the /usage slash command, which opens a dialog on demand. There is no way to passively monitor per-prompt token consumption as you work.

Developers who are cost-conscious or optimizing their prompts have no way to see per-turn token counts inline in the conversation. Having to open /usage after every response is disruptive to the flow.

Proposed Solution

Add a settings option (e.g., showUsagePerPrompt: true in settings.json) that appends a compact usage summary after each assistant response, something like:

↑ 1,234 tokens  ↓ 456 tokens  cache hit: 892  ~$0.004

This could alternatively be surfaced via:

• The statusLine (if the runtime exposes live usage data to status line commands)
• A Stop hook (if usage data is included in the hook's stdin payload)
• A dedicated verbose sub-mode

Alternatives Considered

• statusLine config: The status line can run shell commands, but those commands have no access to the current session's live token usage data.
• Stop hook: Fires after each response, but the hook's stdin payload does not include token usage data.
• viewMode: "verbose": Shows tool call details but not token counts.

Additional Context

This was surfaced as a user request in Claude Code. The /usage dialog already has all the data — it just needs a way to surface it inline or automatically.

🤖 Generated with Claude Code