Claude HUD: The Plugin That Made AI Agents Observable

“I can’t tell if Claude is working or just staring into space.”

If you’ve ever run a complex task through Claude Code, you know the feeling. The terminal goes quiet during a long operation, and there’s no way to know whether context is about to run out, whether the agent is stuck in a loop, or whether anything is happening at all.

An open-source plugin tackled this problem head-on. After using it myself, I wanted to share why it’s worth installing.

What Claude HUD Shows You

Claude HUD places a game-style status bar at the bottom of your terminal. It tracks four things in real time:

• Context usage - A visual progress bar with percentage showing how much of the session’s context window is consumed
• Active tools - Which tools (Read, Edit, Bash, etc.) are currently running
• Sub-agent status - Tracks child agents spawned via Task calls (Opus 4.5, Sonnet, etc.)
• Todo progress - Visualizes completion status of Claude’s internal task list

How It Works Under the Hood

For those curious about the internals, here’s a brief look at what’s happening behind the status bar.

stdin Input Processing

Claude Code passes session state directly as a JSON payload. The plugin calculates context utilization by summing input_tokens, cache_creation_input_tokens, and cache_read_input_tokens, then dividing by context_window_size.

Transcript Parsing

Session logs are processed as a line-by-line stream. The plugin detects tool_use blocks and matches them against tool_result entries to track each tool’s status as running, completed, or errored.

File Path Display

The plugin handles file path display with some thoughtful details:

• Extracts file_path from Read, Write, and Edit tool calls to show which file is being worked on
• Automatically truncates long paths to .../filename.ts format
• Refreshes at roughly 300ms intervals for near-real-time feedback

Why Developers Are Excited

The response from the community has been strong, and the reason is straightforward: this plugin addresses a core pain point of working with AI agents.

• Solves the opacity problem - Long-running tasks no longer leave you guessing whether the agent is stuck, looping, or about to hit the context limit
• Three-line installation - No complex configuration required, just install and go
• Debugging efficiency - In multi-step agent workflows, you can immediately see where things went wrong
• Visual feedback loop - The game-style status bar fits naturally into the vibe-coding culture, making AI-assisted development feel more interactive

A popular comment in the community captured the sentiment well: “Finally I can tell if Claude is working or daydreaming.”

Installation

Setup takes three commands:

/plugin marketplace add jarrodwatts/claude-hud
/plugin install claude-hud
/claude-hud:setup

GitHub: jarrodwatts/claude-hud

Why This Matters

If you use Claude Code regularly, this is worth trying. Working without context anxiety - knowing exactly when a session might run out - provides more peace of mind than you might expect. But the real value is simpler than that: being able to see what Claude is doing at any given moment changes how you work with it.

Observability isn’t just for servers and pipelines. AI agents need it too.