Statusline token counts should include subagent/tool-agent token usage

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

The statusline JSON data exposes context_window.total_input_tokens and context_window.total_output_tokens, but these only reflect the main conversation's token usage. When
Claude Code spawns subagents (Explore, Plan, general-purpose, etc.), their token consumption is not included in these totals.

This makes the token count in custom statuslines misleading — the displayed total can be significantly lower than actual usage.

Proposed Solution

Example

In a session where two Explore agents were running:

Running 2 Explore agents…
├─ Search for TDS scripts · 32 tool uses · 127.3k tokens
└─ Search for TDS helper scripts · 22 tool uses · 118.9k tokens

The statusline showed 35.3k tok — only the main conversation tokens — while the agents alone consumed ~246k tokens. The actual session total should have been ~281k tokens.

Meanwhile, cost.total_cost_usd ($6.34) appeared to correctly include agent costs, suggesting the billing side already aggregates across agents.

Expected behavior

context_window.total_input_tokens and context_window.total_output_tokens (or a new field like total_session_tokens) should include tokens consumed by all subagents, matching
how cost.total_cost_usd already works.

Alternative Solutions

_No response_

Priority

Medium - Would be very helpful

Feature Category

Configuration and settings

Use Case Example

_No response_

Additional Context

_No response_