Skip to content

Dashboard

The OpenJCK dashboard is a live web interface that shows all your agent runs in real-time. No configuration required — it automatically discovers traces as your agents run.

Terminal window
npx openjck

Or with the global install:

Terminal window
openjck ui

The dashboard opens at http://localhost:7823.

The main view shows all agent runs across all agents:

  • Stat row: Total runs, total cost, success rate, average duration
  • Runs table: Every run with status, name, steps, cost, duration, and time
  • Live updates: New runs appear automatically every 2 seconds
  • Time filters: Switch between last 24h, 7 days, or all time

Click any agent name to see runs filtered to that specific agent:

  • Agent-specific stats (runs, cost, success rate, avg steps)
  • Patterns section showing:
    • Most common failure step
    • Most expensive step on average
  • Filtered runs table

Click any run row to see the full execution timeline:

  • Step-by-step timeline with color-coded dots
  • Step detail panel with input/output/error
  • Intelligence panel (for failed runs) showing root cause analysis
  • Click any step dot to inspect that step

The dashboard polls the server every 2 seconds. New runs appear with a brief amber flash animation. The “last updated” counter in the top-right shows how fresh the data is.

Three time ranges are available:

  • 24h: Last 24 hours (default)
  • 7d: Last 7 days
  • all: All time

Click any filter button to switch. The stat row and runs table update immediately.

Each run shows a colored status dot:

  • 🟢 Green: Completed successfully
  • 🔴 Red: Failed with error
  • 🟡 Amber (pulsing): Currently running

Failed runs show an inline intelligence preview:

🔍 Root cause: Step 3 — FileNotFoundError: data/missing.csv
[View Step 3 →]

Click “View Step N” to open the trace detail focused on the root cause step.

The dashboard is responsive:

  • Desktop: Full 4-column stat cards, all table columns
  • Tablet: 2-column stats, all table columns
  • Mobile: 2-column stats, simplified table (hides cost/duration columns)
  • Click any row to open trace detail
  • Click agent name to filter by agent
  • Click “Back” button to return to previous view
  • URL updates automatically — share /trace/:id links

Dashboard shows “Cannot connect to OpenJCK server”

The server isn’t running. Start it with:

Terminal window
npx openjck

No traces appear

Run your instrumented agent first:

from openjck import trace
@trace(name="my_agent")
def run():
# your agent code
pass
run()

Traces appear automatically in the dashboard.