[FEATURE] Surface CLI status line data in Claude app during remote control sessions

Resolved 💬 3 comments Opened Feb 28, 2026 by haighd Closed Mar 4, 2026

Problem

When using Claude Code via remote control (/remote-control or /rc) from the Claude app (claude.ai/code or mobile), the CLI status line information is completely invisible to the user. In a local terminal session, the status line shows critical operational data — context window usage, model info, active tools, etc. — but none of this is surfaced in the remote app UI.

This is especially impactful because:

  1. The user has no visibility into context window consumption. In long agentic sessions (multi-agent sprints, review loops, etc.), context fills up silently. The user can't proactively checkpoint and clear before hitting limits.
  1. The model itself cannot access this data either. The /context CLI command works locally but is sent as plain text in remote sessions (related: #28351). There's no tool or system context that exposes token usage to the model, so the agent can't self-manage session lifecycle.
  1. Remote sessions are where this matters most. Users on remote sessions tend to run longer, more autonomous workflows — exactly the scenario where context awareness is critical.

Proposed Solution

Surface the CLI status line data in the Claude app UI during remote control sessions. This could include:

  • Context window usage (tokens used / remaining, percentage)
  • Model identifier (e.g., claude-sonnet-4-6)
  • Session duration
  • Active background tasks count

Implementation options (non-exhaustive):

  • A persistent status bar in the Claude app remote session UI
  • A collapsible info panel
  • Periodic system-level annotations in the conversation stream

Bonus: Expose to the model

In addition to displaying in the UI, consider making context usage metrics available to the model via system context or a callable tool. This would allow the agent to proactively recommend checkpoints or context clears — useful for autonomous workflows where the user may not be actively monitoring.

Related Issues

  • #28351 — Slash commands (including /context) not supported in remote control sessions
  • #15247 — Built-in context usage indicator in CLI status bar
  • #10828 — Real-time token usage display in CLI status bar
  • #24145 — Bidirectional session sync between CLI and Web UI
  • #21265 — Status line support for VSCode Extension

Environment

  • Claude Code version: 2.1.x
  • Platform: macOS (remote host) → claude.ai/code (client)
  • Use case: Long-running agentic sessions with multi-agent orchestration

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗