[Feature Request] Claude Code for Desktop: native context window usage indicator
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
This is specifically about Claude Code for Desktop (the GUI app), NOT the CLI.
Claude Code for Desktop has no way to see context window usage. The statusLine feature in settings.json only renders in the CLI/terminal version. Desktop users have zero visibility into context consumption until auto-compaction happens unexpectedly.
This causes:
- Surprise compaction mid-task, losing conversation context
- No ability to plan work around context limits
- No way to decide when to start a fresh session vs. continue
Proposed Solution
Add a native context usage indicator somewhere in the Desktop app UI (title bar, footer, sidebar, or settings panel) showing:
- Current context usage percentage (e.g., "72% context used")
- Visual progress bar or meter
- Warning state when approaching compaction threshold
Why This is Not a Duplicate of CLI Issues
Previous issue #20041 was incorrectly auto-closed as a duplicate of CLI-only issues (#7111, #18705, #17431). Those all relate to the terminal statusLine feature, which does not apply to the Desktop app. The Desktop app is a GUI — it has no terminal status bar to render into. This needs a native UI solution.
Alternative Solutions
- The CLI
statusLineconfig insettings.jsonis silently ignored by Desktop - No third-party tools can read Desktop session context data
- Users can ask Claude to estimate usage, but this is imprecise and consumes context itself
Priority
Medium — Would significantly improve session management for Desktop users
Feature Category
Interactive mode (TUI)
Additional Context
The data already exists internally (context_window.used_percentage, context_window.context_window_size) — it just needs a rendering surface in the Desktop UI.
Ref: #20041 (incorrectly closed as duplicate)
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗