GC
User Guide / DashboardMenu
5 min read·Updated 2026-04-10

Dashboard

The dashboard is the main control surface: Chat, Settings, Scheduler, Sessions, Logs, Skills, Diagnostics, and the Plugin Marketplace.

What the dashboard is

The dashboard is the main control surface for the runtime. Every feature has a dedicated page; every runtime action the agent takes is observable through a matching inspection page. This page is a map — what each section is for and where to look for what.

On first boot, before any LLM provider is configured, the runtime shows an onboarding screen that walks you through adding a provider and assigning a model tier. Once that is done, the onboarding screen disappears and the full dashboard becomes available.

Dashboard pages

  • Chat — the main workspace for talking to the agent. Turns stream in real time; tool calls and state changes are visible inline.
  • Settings — runtime and integration options organized by function (see below).
  • Scheduler — cron schedules for Auto Mode goals and tasks.
  • Sessions — historical record of every agent turn with full tool-call detail.
  • Logs — live runtime output. Use for real-time debugging.
  • Skills — browse installed skills, activate them, manage custom skills on disk.
  • Diagnostics — health tools, connectivity checks, memory and tool integrity.
  • Plugin Marketplace — browse and install first-party and community plugins.

Settings catalog

Settings is the biggest surface. It is organized so that related runtime concerns live together rather than in one flat list.

Runtime controls

General settings, LLM providers, model catalog, model router. The core dials every other feature depends on.

Core

Plugins

Plugin marketplace plus plugin-contributed settings pages (one per installed plugin).

Extensions

Runtime and tools

Filesystem, shell, automation, voice, skills, memory, MCP, usage, updates. Feature-level configuration.

Features

Operator safeguards

Rate limiting, security, compaction, diagnostics, SelfEvolving controls. Safeguards and advanced knobs.

Advanced

Authentication

Authentication is intentionally simple: one admin account stored in preferences/admin.json. The first run prints a temporary password to the logs unless an operator-provided password is injected through environment variables or Spring properties.

Authentication model

  • Username is always admin.
  • The generated password is temporary. Rotate it immediately in Settings → Security.
  • BOT_DASHBOARD_ADMIN_PASSWORD can preseed a known password at container start.

Progress and inspection

Chat is backed by a WebSocket transport and emits both assistant output and runtime progress. You can see when the model starts, when tools run, when retries happen, and when a turn finishes or fails.

The same inspection model applies across Sessions (historical context), Logs (live runtime output), and Diagnostics (health tooling). If something looks wrong in Chat, one of the three inspection pages has the detail.

What to do next