TUI Workbench — Terminal Agent Dashboard

VulpineOS ships a terminal-based workbench built with Bubbletea and Lipgloss . It provides a 3-column agent-centric interface for managing the browser kernel, spawning agents, and monitoring conversations in real-time.

Layout


┌─────────────┬──────────────────────┬──────────────┐
│ System Info  │                      │ Agent Detail  │
│ ─────────── │                      │ ────────────  │
│ Kernel: ● ON│    Conversation      │ Name: Scout   │
│ Mem: ██░ 45%│                      │ Status: Active│
│ Risk: Low   │  Agent: Show me the  │ Tokens: 12.4k │
│ Pool: 3/20  │  latest HN posts     │ Proxy: US-1   │
│             │                      │               │
│ Agents      │  VulpineOS: Found    │ Contexts      │
│ ─────────── │  15 stories on the   │ ────────────  │
│ ● Scout     │  front page...       │ ctx-a8f2      │
│ ◌ Crawler   │                      │  hn.com       │
│ ✓ Archiver  │  [input area]        │               │
│ ⏸ Monitor   │                      │               │
└─────────────┴──────────────────────┴──────────────┘

Left Column

System Info — kernel status, memory usage, detection risk gauge, context pool utilization, page count
Agent List — all agents with status icons: ● active, ◌ idle, ✓ completed, ⏸ paused, ✗ failed

Center Column

Conversation — full-height scrollable chat display. Messages are color-coded by role (user, assistant, system). Text input at the bottom, Claude CLI style.

Right Column

Agent Detail — name, status, token usage, task description, fingerprint profile, assigned proxy
Context List — active browser contexts for the selected agent with URLs

Agent Lifecycle


CREATE ──→ ACTIVE ──→ PAUSED ──→ ACTIVE ──→ COMPLETED
                                              or FAILED

Agents are persistent — their conversations are stored in the SQLite vault and survive restarts. Each agent gets a deterministic fingerprint generated from a seed, so resuming an agent maintains the same browser identity.

Key Bindings

Key	Action
`n`	Create new agent (prompts for name, then task)
`j` / `k`	Navigate agent list up/down
`Enter`	Send message to selected agent
`p`	Pause selected agent (`/savestate` sent to OpenClaw)
`r`	Resume paused agent with saved session
`x`	Delete agent and its conversation history
`Tab`	Cycle focus: Agent List → Conversation → Agent Detail → Contexts
`Esc`	Cancel current input / go back
`Shift+S`	Open settings panel (replaces center column)
`c`	Re-open setup wizard to change provider/model
`q`	Quit VulpineOS

Settings Panel

Press Shift+S to open the 3-tab settings panel:

General — AI provider, model, API key
Proxies — import proxy URLs, test latency, resolve geo, delete
Skills — toggle global OpenClaw skills on/off

Theme

The TUI uses a violet/cyan/green/amber/red palette defined in internal/tui/shared/styles.go, matching the VulpineOS brand.