Feature Request: Real-time token usage / output TPS indicator during generation (similar to Claude Code)

[wtq524176336-source]

1. The Problem (Pain Point)

Currently, when using Droid CLI, it feels like a "black box" during long generations or tool executions. There is no real-time feedback on how many tokens are being consumed or generated.

Users cannot easily tell if:

• The model is actively thinking/generating in the background.
• The CLI is stuck in an infinite tool-call loop.
• The network or API is hanging.

While commands like /cost and /context are helpful, they only provide "post-execution" statistics. The showTokenUsageIndicator setting also does not provide a live, ticking counter of output tokens.

2. Expected Behavior & Solution

Please consider adding a live token usage tracker or a Tokens Per Second (TPS) meter that updates dynamically while the model is streaming its response.

Inspiration (Claude Code):
In Claude Code, users or third-party extensions can monitor live token increments (e.g., 10... 50... 120 tokens) and cost during generation.

It would be great if Droid CLI could:

1. Provide a native, real-time ticking token counter in the terminal UI (or status bar).
2. OR expose variables like total_output_tokens and current_cost to the /statusline configuration so we can build our own live progress bars.
3. OR output a continuous local log file (e.g., .jsonl) during generation that third-party trackers can listen to.

3. Why This Matters

• Reduces user anxiety: We immediately know the CLI hasn't frozen.
• Cost control: Users can instinctively hit Ctrl+C if they see tokens exploding out of control.
• Better Developer Experience: Matches the transparency of competing tools like Claude Code.