Claude Session Dashboard

Claude Session Dashboard

See exactly where your context window goes. Trace every agent delegation. Understand how your Claude Code workflows actually execute — all from your local ~/.claude, no data sent anywhere.

Why?

If you run Claude Code with agentic workflows and custom skills, you've probably felt this: sessions that start sharp and gradually get worse. Claude repeating itself, missing earlier context, giving shallower answers. The context window is filling up silently and you have no way to see it.

The fix is agent delegation — each subagent starts with a fresh context, keeping your main session lean. But most people do this by instinct, not by design. You can't optimize what you can't see.

This dashboard makes it visible. It reads your local session files, parses the JSONL logs, and shows you what's actually happening inside your workflows — which agents were dispatched, in what order, how much context each step consumed, and where your tokens are going.

Everything runs entirely on your machine. Read-only — it never modifies any Claude Code data.

Features

Agent delegation timeline ← the unique one
- Gantt-style diagram showing every subagent dispatch in a session, in sequence
- See the full delegation chain: which agents ran, in what order, how long each took
- Token usage per agent — understand exactly how each delegation affects context consumption
- Zoom controls for detailed inspection of complex multi-agent workflows

Context window visualization
- Live breakdown of how your 200K context window is being used: input, output, cache reads, cache writes
- See at a glance whether your main session is staying lean or accumulating context
- Per-session and per-agent cost estimates with per-model and per-category breakdowns

Session browsing and search
-…