🐾 Agents in a Box — Masterclass Resource

ClawBuddy LiteKit

Build your own AI Agent Command Center with this Lovable prompt. A dark-themed, glass-morphic dashboard that serves as mission control for your AI workforce.

A Full Command Center in One Prompt

Paste this prompt into Lovable and get a working AI agent dashboard in minutes. Five tabs, dark theme, glass-morphic cards, all running on local mock data — ready to connect to your real agents.

📊

Command Deck

Metric cards, activity feed, and agent status panels with live-looking updates

🤖

Agent Profiles

Profile cards for each AI agent showing role, skills, and performance stats

📋

Kanban Board

4-column task board with draggable cards, priority badges, and agent assignments

📝

AI Log

Chronological feed of agent observations, decisions, and milestones with category filters

⚖️

Agent Council

Structured debate sessions where agents discuss decisions with turn limits

🎙️

Meeting Intelligence

KPI cards, type distribution chart, searchable feed, expandable detail panels with action items

Premium Design

Space Grotesk + JetBrains Mono, emerald accents, Framer Motion animations, responsive

💡 How to Use This

Copy the prompt below and paste it into Lovable (or any AI code builder). It generates a complete React app with shadcn/ui, Framer Motion, and mock data. Once built, you can swap the mock data for real ClawBuddy API calls to connect your agents.

ClawBuddy LiteKit — Lovable Prompt

This is the full prompt. Click "Copy" to grab it, then paste into Lovable.

lovable-prompt.md
# Lovable Prompt: ClawBuddy LiteKit — AI Agent Command Center ## What to Build Build **ClawBuddy** — a sleek, dark-themed AI agent command center dashboard. This is a standalone React app that serves as mission control for AI agents. Think of it as a "NASA mission control" for your AI workforce. It should feel premium, futuristic, and data-rich. This is a **demo/starter kit** — all data is mock/local state. No backend needed. The dashboard should feel alive with animations, glowing accents, and real-time-looking updates. --- ## Design System ``` Background: #0a0a0f (near-black with slight blue tint) Surface cards: rgba(17, 24, 39, 0.7) with backdrop-filter: blur(16px) Surface border: rgba(255, 255, 255, 0.06) Primary accent: #10b981 (Emerald green — the signature color) Cyan accent: #06b6d4 (for "live" and "active" states) Amber accent: #f59e0b (warnings, pending) Red accent: #ef4444 (errors, alerts) Text primary: #f9fafb Text secondary: #9ca3af Text muted: #6b7280 Font (headings): Space Grotesk (import from Google Fonts) Font (mono/data): JetBrains Mono (import from Google Fonts) Glow effects: CSS box-shadow with primary/cyan at 20-40% opacity Card style: Glass-morphic — use a reusable "glass-card" CSS class: background: rgba(17, 24, 39, 0.7); backdrop-filter: blur(16px); border: 1px solid rgba(255, 255, 255, 0.06); border-radius: 12px; Animations: Framer Motion for all transitions (stagger, fade, slide) Icons: Lucide React UI components: shadcn/ui (Badge, Tabs, Select, ScrollArea, Dialog, Button, Input, Toggle) Charts: Recharts (PieChart, BarChart) — used in Meeting Intelligence tab Sanitization: DOMPurify — sanitize markdown summaries in meeting details Date formatting: date-fns (format, parseISO, isAfter, subDays) ``` --- ## Page Layout Single page app with a **top header** and **tabbed content area**: ``` ┌──────────────────────────────────────────────────────────────────────────┐ │ 🐾 ClawBuddy 🟢 Agent Alpha: Online [⚙️] │ │ AI Agent Command Center Last seen: just now │ ├──────────────────────────────────────────────────────────────────────────┤ │ [Command Deck] [Agents] [Task Board] [AI Log] [Council] [Meetings] │ ├──────────────────────────────────────────────────────────────────────────┤ │ │ │ (Tab content area) │ │ │ └──────────────────────────────────────────────────────────────────────────┘ ``` --- ## TAB 1: Command Deck (Dashboard Overview) The landing tab. A grid of metric cards + recent activity. ### Row 1: Metric Cards (4 cards, horizontal) Each card: glass-card, icon with emerald glow, large number, small trend indicator. ### Row 2: Recent Activity Feed (Left 60%) + Agent Status (Right 40%) **Activity Feed:** Scrollable list of recent events with agent emoji, action text, relative timestamp. Newest at top. Glass-card container. **Agent Status Panel:** Small cards for each agent with colored status dot (green=active, yellow=idle, red=error, gray=offline), agent name, current activity, last seen. --- ## TAB 2: Agent Profiles Grid of agent profile cards (3 columns desktop, 1 mobile). Each card shows emoji, name, subtitle, type, role, status, tasks completed, accuracy, skills list, and a "View Details" button. Three agents pre-populated: | Agent | Emoji | Type | Role | Accent Color | |-------|-------|------|------|-------------| | Agent Alpha | 🤖 | Code Agent | Lead Engineer | Emerald #10b981 | | Dispatch Bot | 📋 | Coordinator | Operations Director | Amber #f59e0b | | Audit Bot | 🛡️ | Quality Agent | Compliance Officer | Cyan #06b6d4 | Users can customize these agent names and roles to match their own AI workforce. --- ## TAB 3: Task Board (Kanban) A 4-column kanban board: To Do, Doing, Needs Input, Done. Task cards: glass-card, title, assigned agent emoji, progress percentage (if doing), priority badge (low/medium/high/urgent as colored dots). Cards should be draggable between columns. Pre-populate with 10 mock tasks spread across all 4 columns. --- ## TAB 4: AI Log A chronological feed of agent log entries with category badges: `observation` (emerald), `general` (gray), `reminder` (amber), `fyi` (cyan). Filter dropdown to show one category at a time. Pre-populate with 8-10 log entries from different agents. --- ## TAB 5: Council Structured debate sessions between agents. Show 2 example council sessions. Each session: collapsible card with question, participant chips (emoji + name + sent/limit + status icon), status badge, message feed when expanded. Messages show agent emoji, name, message number (#1, #2, #3), and relative timestamp. --- ## TAB 6: Meeting Intelligence A full meeting tracking dashboard. Pre-populate with 8-10 mock meetings. ### Row 1: KPI Cards (4 cards, horizontal) Each card: glass-card style, icon with emerald glow, large number. - **Total Meetings** — Calendar icon, number (e.g. 247) - **This Week** — TrendingUp icon, number (e.g. 8) - **Open Action Items** — CheckSquare icon, number (e.g. 12) - **Avg Duration** — Clock icon, formatted (e.g. "34m") ### Row 2: Charts (2 columns) - **Left: Meeting Type Distribution** — Recharts donut/pie chart with 6+ meeting types, each with a distinct color: 1-on-1 (#60a5fa), external (#a78bfa), sales (#34d399), team (#fb923c), standup (#818cf8), planning (#2dd4bf). Legend below. - **Right: Monthly Trend** — Recharts bar chart. Group meetings by month, emerald bars. ### Row 3: Search + Filters - Full-width search input with Search icon — filters meetings by title - Filter pills: meeting type multi-select, date range (7d/30d/90d/All), "Has Action Items" toggle, "External Only" toggle - Sort dropdown: Most Recent, Oldest First, Longest Duration ### Row 4: Meeting Feed Scrollable list of meeting cards (paginated, 25 per page): **Each card shows:** - Meeting type badge (colored pill) - Title (bold) - Date (relative or absolute) - Duration ("45m" or "1h 15m") - Attendee avatars (circular initials, max 3 shown + "+N" overflow) - Action item count badge (if any) - Globe icon if has external participants **Click to expand** — slides open detail panel: - Full summary text (mock markdown) - Action items list with checkboxes - AI insights line (muted, Sparkles icon) - Attendee list, external domains - "Open Recording" and "Share Link" buttons - "Send To..." dropdown: Action Items, Proposals, Lead Magnets ### Mock Data Shape ```json { "type": "meeting", "title": "Weekly Standup with Engineering", "date": "2026-02-25T10:00:00Z", "duration_minutes": 30, "duration_display": "30m", "attendees": ["Alice", "Bob", "Charlie"], "summary": "Discussed sprint progress. Backend API 80% complete...", "action_items": [ {"task": "Review PR #42", "assignee": "Alice", "done": false}, {"task": "Update docs", "assignee": "Bob", "done": true} ], "ai_insights": "30 min meeting with 3 attendees", "meeting_type": "standup", "sentiment": "positive", "has_external_participants": false, "external_domains": [], "fathom_url": null, "share_url": null } ``` Pre-populate with diverse meeting types: 2 standups, 2 sales calls (with external attendees), 1 interview, 1 all-hands, 2 1-on-1s, 1 planning session, 1 team meeting. --- ## Header Component Left: Paw icon (🐾) + "ClawBuddy" in bold + subtitle "AI Agent Command Center" Right: Active agent status — colored dot + agent name + current activity + last seen Glass-card background with subtle emerald glow on left border. The status dot should pulse gently when the agent is active. --- ## Animations & Polish 1. Page load: Tabs fade in with stagger (50ms delay between each) 2. Tab switch: Content crossfades (200ms) 3. Cards: Hover effect — subtle lift + border glow intensifies 4. Status dots: Pulse animation for active agents (2s loop) 5. Metric cards: Numbers count up on first render 6. Log entries: Slide in from left with stagger 7. Task cards: Subtle scale on hover (1.02x) 8. Council expand: Smooth height animation with message stagger 9. Unread badge: Gentle bounce when count > 0 --- ## Responsive - Desktop (>1024px): Full layout, 3-column agent grid, side-by-side dashboard panels - Tablet (768-1024px): 2-column agent grid, stack dashboard panels vertically - Mobile (<768px): Single column everything, kanban columns horizontal scroll --- ## Important Notes 1. No backend. Everything runs on local mock data and React state. 2. The vibe is "premium dark dashboard." Think GitHub Dark, Linear, Vercel. 3. Glass-morphism is the signature look. Every card gets backdrop blur. 4. Emerald (#10b981) is THE accent color. Cyan is secondary. 5. Performance: Keep animations subtle and GPU-friendly. 6. Branding: The app is called "ClawBuddy" with the paw icon (🐾).

⚠️ After Building

The LiteKit generates with mock data. To connect it to real agents, you'll need to set up the ClawBuddy backend (Supabase edge functions) and replace the mock data hooks with real API calls. The other three prompts on this site (Cognitive Memory, Self-Evolution, 8-Phase Pipeline) teach your agents HOW to talk to the dashboard.

🛠️ Starter vs Production

This prompt creates a V1 scaffold. Production requires wiring real data sources, QA passes, error handling, retry logic, and safety gates. Treat the output as a working prototype — not a ship-ready system.