office-cam – OpenClaw Skill

---

name: office-cam description: Multi-camera system for office/home monitoring. Supports USB webcams (Logitech), WiFi Wyze cameras (RTSP), and ESP32 cameras. Use to check rooms, capture photos on demand, or monitor multiple spaces. metadata: clawdbot: emoji: 🎥

Office Cam - Multi-Camera System

Whole-house camera network supporting multiple camera types.

Supported Cameras

Type	Reliability	Setup	Best For
🖥️ USB Webcam	⭐⭐⭐⭐⭐ Instant	Plug & play	Desk, office
📹 Wyze PTZ/v3	⭐⭐⭐⭐⭐ Reliable	RTSP stream	Rooms, garage, shed
📡 ESP32-CAM	⭐⭐⭐ Experimental	ESP-NOW wireless	DIY, battery-powered

Quick Start

# USB Webcam (instant)
~/clawd/skills/office-cam/scripts/capture.sh

# Wyze PTZ (one-time setup required)
~/clawd/skills/office-cam/scripts/wyze-dashboard add
~/clawd/skills/office-cam/scripts/wyze-dashboard capture shed

# ESP32-CAM (ESP-NOW wireless)
# Flash firmware, auto-discovers base station, sends photos wirelessly

---

🖥️ USB Webcam (Logitech)

Most reliable - works instantly.

Basic Capture

cd ~/clawd/skills/office-cam
./scripts/capture.sh /tmp/office.jpg

Motion Detection ⭐ NEW

# Start motion detection (runs continuously)
./scripts/motion-detect.sh

# Captures saved to ~/.clawdbot/motion-captures/
# Automatically triggers when motion detected
# 5-second cooldown between captures

🔥 OVERWATCH Mode (AI Monitoring) ⭐⭐⭐

# Start AI-powered continuous monitoring
./scripts/overwatch start

# Check status
./scripts/overwatch status

# Stop monitoring
./scripts/overwatch stop

What Overwatch does:

• 👀 Monitors webcam 24/7 in background
• 🚨 Detects motion and saves alert
• 🔔 Can trigger notifications (integrate with OpenClaw)
• 💾 Saves motion captures to ~/.clawdbot/overwatch/

Or say: "Start overwatch" or "Keep overwatch"

🤖 SMART OVERWATCH (AI-Escalated) ⭐⭐⭐⭐⭐

# Start smart monitoring (zero cost until trigger)
./scripts/smart-overwatch start

# I will check for triggers and analyze when found

How it works:

1. 👀 Local motion detection (file size comparison, zero API cost, runs always)
2. 🚨 Motion detected → creates trigger file in ~/.clawdbot/overwatch/triggers/
3. 🤖 I detect the trigger → analyze image with vision model (only when motion happens)
4. 👤 Person found? → Send alert & optionally continue watching
5. 📊 No person? → Delete trigger, back to local monitoring

What you can say:

• "Start smart overwatch"
• "Keep watch and tell me if you see anyone"
• "Monitor for people only"
• "Check overwatch triggers" (to manually check for pending analysis)

Cost: Zero until motion detected, then only when I analyze the image

📸 Instant Telegram Photo ⭐⭐⭐⭐⭐

Quick capture and send - the simplest way to check your camera

Just say:

• "Show me the office"
• "What's on camera?"
• "Send me a photo"

What happens:

1. I instantly capture from your webcam
2. Analyze for clear faces/activity
3. Send best image directly to your Telegram

No stream needed - just instant photos when you want them.

🔥 OVERWATCH PRO (Full Control System) ⭐⭐⭐⭐⭐⭐

The complete solution - Telegram alerts, live stream, remote control

# Start the full system
./scripts/overwatch-pro start

# Setup Telegram (one-time)
./scripts/overwatch-pro setup

# Get live stream URLs
./scripts/overwatch-pro stream

# Take manual photo
./scripts/overwatch-pro capture

# Check status
./scripts/overwatch-pro status

# View live log
./scripts/overwatch-pro log

# Stop
./scripts/overwatch-pro stop

Features:

• 🚨 Instant Telegram alerts when motion detected
• 🌐 Live MJPEG stream at http://localhost:8080
• 📱 Auto-refreshes every 2 seconds
• 📸 Saves all captures to ~/.clawdbot/overwatch/
• 🤖 Telegram bot replies - respond to alerts with commands
• 🌅 Morning report (8 AM daily via cron)

Telegram Commands (reply to motion alert):

• analyze - I'll check the image and tell you what I see
• stream - I'll send you the live stream link
• capture - I'll take a fresh photo right now

Network Access:

# Local
http://localhost:8080

# From another device on same network
http://$(hostname -I | awk '{print $1}'):8080

Morning Report (with Photo):

• Auto-generated at 8 AM (America/Denver timezone)
• Sends a fresh morning photo of your office
• Summarizes overnight motion events
• Offers to analyze captures
• Provides quick command reference

Requirements for motion detection:

brew install imagemagick  # For image comparison

---

📹 Wyze Cameras

One-Time Setup:

1. Enable RTSP in Wyze App:

   • Open Wyze app → Camera Settings → Advanced Settings
   • Enable RTSP, set password
   • Copy the RTSP URL

2. Add to system:

   ~/clawd/skills/office-cam/scripts/wyze-dashboard add
   # Enter camera name (e.g., "shed")
   # Enter RTSP URL
   

3. Capture:

   # Single camera
   ~/clawd/skills/office-cam/scripts/wyze-dashboard capture shed
   
   # All cameras
   ~/clawd/skills/office-cam/scripts/wyze-dashboard capture-all
   
   # Quick one-liner (after setup)
   ~/clawd/skills/office-cam/scripts/wyze-capture shed
   

---

📡 ESP32-CAM (ESP-NOW Wireless)

No WiFi router needed! Direct wireless between ESP32s.

Architecture:

┌─────────────────┐      ESP-NOW      ┌─────────────────┐
│  ESP32-CAM      │  ═══════════════► │  ESP32 Base     │◄── USB ──► Mac
│  (Camera Node)  │   (100m range)    │  (Receiver)     │
└─────────────────┘                   └─────────────────┘
        Battery powered                  Plugged into computer

Flashing:

Base Station (plain ESP32, plugged into Mac):

cd ~/clawd/skills/office-cam/firmware/espnow-base
pio run --target upload

Camera Node (ESP32-CAM with camera):

cd ~/clawd/skills/office-cam/firmware/espnow-cam-auto
pio run --target upload

How it works:

1. Base station broadcasts "beacon" every 2 seconds
2. Camera auto-discovers base station
3. Camera captures and sends photo wirelessly
4. Base receives and saves to SD/outputs to serial

Range: 100+ meters (no WiFi needed!)

---

Files

Script	Purpose
capture.sh	USB webcam
wyze-capture	Quick Wyze capture
wyze-dashboard	Multi-cam management
firmware/espnow-base/	ESP32 receiver
firmware/espnow-cam-auto/	ESP32-CAM transmitter

---

Troubleshooting

Wyze "Connection refused":

• Verify RTSP enabled in Wyze app
• Check username/password
• Ensure same WiFi network

ESP32-CAM not connecting:

• Keep within 10 feet for testing
• Check LED: blinking = searching, solid = capturing
• Try resetting both boards

USB webcam not working:

• brew install imagesnap
• System Settings → Privacy → Camera

📹 ClawCamera — Multi-Camera Surveillance with AI

Professional-grade office/home monitoring system with motion detection, continuous monitoring (Overwatch), and AI-powered analysis. Built for OpenClaw.

🎯 What It Does

✅ Instant snapshots — Ask "Is anyone here?" and get AI-powered visual answers

✅ Motion detection — Alerts when movement is detected (configurable modes)

✅ Overwatch mode — 24/7 background monitoring with periodic check-ins

✅ Smart Overwatch — Local motion detection (zero cost) → AI escalation only when needed

✅ Multi-camera — USB webcams, Wyze RTSP, ESP32-CAM all supported

✅ Telegram integration — Instant notifications with images

✅ No secrets in git — Comprehensive .gitignore + env-based config

🚀 Quick Start

1. Install

git clone https://github.com/Snail3D/ClawCamera.git
cd ClawCamera
npm install

2. Configure

# Copy example config
cp .env.example .env

# Add your Telegram credentials for notifications
export TELEGRAM_TOKEN=your_bot_token
export TELEGRAM_CHAT_ID=your_chat_id

3. Use

One-shot capture:

./scripts/capture.sh
# → Captures photo
# → Output: Saved to /tmp/capture.jpg

Start Overwatch:

./scripts/overwatch-pro start
# → Continuous background monitoring
# → Motion alerts sent to Telegram
# → Live stream at http://localhost:8080

Smart Overwatch (AI-escalated):

./scripts/smart-overwatch start
# → Local motion detection (zero API cost)
# → Creates triggers for AI analysis
# → AI only runs when motion detected

📚 Documentation

• SKILL.md — Full feature list, configuration, and integration guide (for OpenClaw Hub)
• guides/esp32-setup.md — ESP32-CAM firmware & deployment
• guides/wyze-setup.md — Wyze camera RTSP configuration
• guides/troubleshooting.md — Common issues & solutions

🏗️ Architecture

┌─────────────────────────────────────────┐
│  User Request (Chat / Voice Command)    │
└────────────────┬────────────────────────┘
                 │
        ┌────────▼────────┐
        │    Capture      │
        │       OR        │
        │  Motion Check   │
        │       OR        │
        │    Overwatch    │
        └────────┬────────┘
                 │
        ┌────────▼────────────────────┐
        │      Image Storage          │
        │   ~/.clawdbot/overwatch/    │
        └────────┬────────────────────┘
                 │
        ┌────────▼──────────────────┐
        │   OpenClaw AI Analysis    │
        │   (On-demand via triggers)│
        └────────┬──────────────────┘
                 │
        ┌────────▼──────────────────┐
        │   Telegram Notification   │
        │      Image + Alert        │
        └────────────────────────────┘

How AI Analysis Works:

1. Local motion detection runs continuously using file size comparison (zero API cost)
2. When motion is detected, a trigger file is created in ~/.clawdbot/overwatch/triggers/
3. OpenClaw detects the trigger and analyzes the image using its configured vision model
4. If a person is detected, notifications are sent via Telegram
5. If no person, monitoring continues silently

🎯 Monitoring Modes

🔴 Overwatch Pro (Full-Featured)

./scripts/overwatch-pro start

• 🚨 Instant Telegram alerts on motion
• 🌐 Live MJPEG stream at http://localhost:8080
• 📱 Remote commands via Telegram replies
• 💾 Saves all captures to ~/.clawdbot/overwatch/

Telegram Commands (reply to motion alert):

• analyze — Request AI analysis of the image
• stream — Get live stream link
• capture — Take a fresh photo

🟡 Smart Overwatch (Cost-Optimized)

./scripts/smart-overwatch start

• 👀 Local motion detection (zero API cost, runs always)
• 🚨 Motion detected → creates trigger file
• 🤖 AI analyzes only when trigger exists
• 👤 Person found? → AI starts continuous watching
• 📊 No person? → Back to local monitoring

🔵 One-Shot Capture

./scripts/capture.sh

Instant photo capture on demand. Great for quick checks.

📸 Multi-Camera Support

USB Webcam (Instant)

Plug in any USB webcam and capture immediately.

./scripts/capture.sh --device /dev/video0

Requirements:

brew install imagesnap  # macOS

Wyze Camera (RTSP)

Stream from Wyze PTZ or v3 cameras over your local network.

export WYZE_IP=192.168.1.100
./scripts/capture.sh --device wyze

Setup:

1. Enable RTSP in Wyze app → Camera Settings → Advanced Settings
2. Set RTSP password
3. Use the provided RTSP URL

ESP32-CAM (Wireless)

Deploy an ESP32-CAM to remote locations with OV2640 sensor + WiFi.

# See guides/esp32-setup.md for full firmware & config
./scripts/capture-esp32.sh --ip 192.168.1.50

🔐 Security & Privacy

✅ No Secrets in Git

• Comprehensive .gitignore blocks all sensitive files
• API keys stored in .env (never committed)
• Credentials in credentials.h or config.json are ignored
• auth.json, secrets/ folder all blocked

✅ Local-First Architecture

• Motion detection runs locally (no cloud calls)
• Images stored in ~/.clawdbot/overwatch/ (local filesystem)
• AI analysis only happens when triggers are detected
• No continuous API usage or streaming to cloud

✅ Configurable Data Retention

Captures are stored locally. Add a cron job for cleanup:

# Delete captures older than 7 days
find ~/.clawdbot/overwatch -name "*.jpg" -mtime +7 -delete

🛠️ Requirements

• macOS or Linux
• imagesnap (brew install imagesnap) for USB webcams
• imagemagick (brew install imagemagick) for motion detection
• Python 3 for Overwatch scripts
• Telegram Bot Token for notifications (optional)

📁 File Structure

ClawCamera/
├── scripts/
│   ├── capture.sh           # One-shot USB webcam capture
│   ├── overwatch-pro        # Full monitoring with Telegram
│   ├── smart-overwatch      # Cost-optimized AI escalation
│   ├── motion-detect.sh     # Basic motion detection
│   ├── wyze-dashboard       # Wyze camera management
│   └── capture-esp32.sh     # ESP32-CAM capture
├── firmware/
│   ├── espnow-base/         # ESP32 receiver firmware
│   └── espnow-cam-auto/     # ESP32-CAM transmitter firmware
├── guides/
│   ├── esp32-setup.md
│   ├── wyze-setup.md
│   └── troubleshooting.md
├── .env.example             # Environment template
├── .gitignore               # Prevents secrets in git
├── SKILL.md                 # OpenClaw integration guide
└── README.md                # This file

🤝 OpenClaw Integration

This skill is designed to work with OpenClaw. When properly configured:

• "Show me the office" → Instant photo + analysis
• "Start overwatch" → Begin monitoring
• "Is anyone there?" → One-shot capture + AI check

See SKILL.md for full OpenClaw integration details.

License

MIT