Most LLM debugging happens blind: you only notice context exhaustion, tool calls, or agent activity after a failure. This plugin makes those signals always visible in Claude Code's statusline so you can spot context saturation, watch tools read/edit/search files, and track agent/todo progress without switching views.
What Sets It Apart
- Token-accurate context and usage: reads native context/usage data from Claude Code instead of estimating, so percentage and token counts match what the model is actually using — useful for long-context and 1M-token sessions.
- Inline, terminal-native UI: implements Claude Code's statusline API (no separate panes or tmux), updating ~300ms for near-real-time visibility during interactive coding or agent runs.
- Activity parsing: extracts tool calls, file edits, grep/read counts and shows which subagents are running and their status, making it easier to debug multi-step agent workflows.
- Configurable granularity: presets (Full/Essential/Minimal) and JSON config let you toggle git, tools, agents, todos, and color thresholds to reduce noise for different workflows.
Who It's For
Great fit if you run Claude Code sessions that hit large context windows, use tools or agent chains frequently, or need lightweight in-terminal observability while developing or debugging conversational agents. It helps developers catch context/usage limits and follow asynchronous agent activity without reading long transcripts.
Look elsewhere if you don't use Claude Code (the plugin is specific to Claude Code v1.0.80+), need GUI-based dashboards outside the terminal, or require centralized multi-user telemetry — this plugin focuses on per-session terminal visibility rather than team analytics.
Where It Fits
Think of it as a statusline-level observability tool for LLM-driven development: complementary to log/telemetry systems and local dashboards, but optimized for immediate, in-context feedback while you work in Claude Code.
