Context usage not visible during remote control sessions

Problem

When using Claude Code via remote control (e.g., mobile app), there is no way to see context window usage percentage.

The terminal-based status line displays context % in the local terminal, but during remote sessions there is no terminal status line — only the chat interface. The model itself has no direct visibility into its own context usage either.

Users who build custom hooks to surface context alerts (e.g., a UserPromptSubmit hook that reads state from the status line) also lose this capability during remote sessions, since the status line process isn't running.

Impact

• Users on mobile/remote cannot gauge when to checkpoint or start a fresh session
• The only signal is when the model starts degrading or the session hits the limit — by then it's too late for a clean handoff
• Power users who rely on context-aware workflows lose a key signal

Suggested approaches

• Expose context usage metadata (tokens used, percentage, limit) as part of the session state accessible to hooks, even when no terminal status line is running
• Surface context % in the remote control UI itself (similar to how "Idle" status is already shown)
• Allow hooks to query context usage directly from the API/runtime rather than depending on the status line process

Environment

• Platform: Windows 11 (Git Bash)
• Remote control via Claude mobile app (iOS)
• Claude Code with 1M context window (Max plan)