Agency HQ

A pixel art office visualization for your AI agent team. Shows real-time agent status, activity feeds, and personality-driven banter. Works with OpenClaw in...

Agency HQ — AI Agent Office

A real-time pixel art visualization of your AI agent team. Agents move between rooms (office, kitchen, game room, server room) based on their actual status. Includes a live activity feed, agent spotlight cards, and personality-driven chat.

When to Use

• You want a visual dashboard showing what your agents are doing
• You want to showcase your agent team to others (demo mode)
• You want a fun, always-on display of your OpenClaw setup

Setup

1. Clone and Install

git clone https://github.com/enjinstudio/agency-hq.git
cd agency-hq
npm install

2. Configure Mode

Copy .env.example to .env.local:

cp .env.example .env.local

Set ARENA_MODE=live to connect to your OpenClaw instance, or leave as demo for simulated data.

3. Customize Your Agents

Edit src/lib/agents.ts. Each agent needs:

Field	Description
id	Must match your OpenClaw agent ID (e.g., main, dev, research)
name	Display name
emoji	Avatar emoji
role	Role label shown in spotlight
model	Model name shown in spotlight
color	Hex color for theme and pixel art
desk	Desk position: command, dev, trading, research, design, security, content, strategy, engineering, pm, finance
accessory	Pixel art accessory: glasses, hat, badge, headphones, scarf, cap, bowtie, visor, antenna, crown, monocle

4. Customize Chat Lines (Optional)

Edit src/lib/agent-chat.ts to write personality-driven banter. Each agent has:

• general — random lines said to the room
• to{AgentName} — directed lines at specific agents (30% chance)

5. Run

# Development
npm run dev

# Production
npm run build && npm start

6. Deploy to Vercel (Optional)

Push to GitHub and import in Vercel. Demo mode activates automatically on Vercel.

How Live Mode Works

In live mode, the app reads from your OpenClaw directory:

• Agent status — scans ~/.openclaw/agents/{id}/sessions/*.jsonl for recent activity
• Activity feed — extracts user messages and tool calls from session files
• System stats — reads /proc/loadavg, runs free -m and df
• Room assignment — active agents → office, idle → kitchen/game room, offline → rest room

No database. No external APIs. Just reads files from disk.

Requirements

• Node.js 20+
• OpenClaw (for live mode only)
• npm

Notes

• The chat system generates fun flavor text — it's not real agent communication
• Demo mode works everywhere, including Vercel, with zero configuration
• The pixel art renderer uses Canvas2D — no WebGL required, works in all browsers
• Mobile responsive — stacks vertically on small screens