Skip to main content

​
Robot Kit Overview

The ZeroClaw Robot Kit is a complete toolkit for building AI-powered robots. Designed for Raspberry Pi deployment with offline Ollama inference, it provides all the tools an AI agent needs to interact with the physical world.

​
Features

ToolDescriptionHardware
driveOmni-directional movement (forward, strafe, rotate)Motor controller, wheels
lookCamera capture + vision model descriptionUSB webcam or Pi Camera
listenSpeech-to-text via Whisper.cppUSB microphone
speakText-to-speech via Piper TTSSpeaker + amplifier
senseLIDAR, motion sensors, ultrasonic distanceRPLidar, PIR, HC-SR04
emoteLED expressions and sound effectsLED matrix, buzzer

​
Architecture

┌─────────────────────────────────────────────────────────┐
│                 ZeroClaw + Ollama                       │
│              (High-Level AI Brain)                      │
└───────────────────────┬─────────────────────────────────┘
                       │
         ┌─────────────┬┼─────────────┐
         │             │             │
    ┌─────┼────┐  ┌──────┼────┐  ┌──────┼────┐
    │ drive   │  │  look    │  │  speak   │
    │ sense   │  │  listen  │  │  emote   │
    └────┬─────┘  └────┬─────┘  └────┬─────┘
         │            │             │
         └────────────┼─────────────┘
                        │
    ┌───────────────────┼────────────────────┐
    │    SafetyMonitor (parallel)              │
    │  • Pre-move obstacle check              │
    │  • Proximity-based speed limiting       │
    │  • Bump sensor response                 │
    │  • Watchdog auto-stop                   │
    │  • Hardware E-stop override             │
    └───────────────────┬────────────────────┘
                        │
                        ▼
    ┌───────────────────────────────────────┐
    │        Hardware Layer               │
    │  Motors, Camera, Mic, Speaker, LEDs    │
    └───────────────────────────────────────┘

​
Hardware Requirements

​
Minimum Setup

  • Computer: Raspberry Pi 4 (4GB) or Pi 5
  • Camera: USB webcam
  • Microphone: USB microphone
  • Speaker: Speaker with amplifier
  • Motors: Motor controller (L298N, TB6612) + 4 DC motors + omni wheels
  • Computer: Raspberry Pi 5 (8GB)
  • Storage: 64GB+ NVMe drive
  • Sensors: RPLidar A1, PIR motion sensors, HC-SR04 ultrasonic
  • Display: LED matrix (8x8) for expressions
  • Safety: E-stop button, bump sensors

​
Full Bill of Materials

ComponentModelPrice (approx)
Raspberry Pi 58GB$80
NVMe SSD64GB+$15-30
Motor ControllerL298N or TB6612FNG$5-15
DC Motors4× TT Motors$20-40
Omni Wheels4× 48mm$10-20
LIDARRPLidar A1$100
CameraPi Camera 3 or USB webcam$25-50
MicrophoneUSB mic$10-30
Speaker3W amp + speaker$10-20
LED Matrix8×8 WS2812B$10
E-Stop ButtonBig red mushroom$5
Bump Sensors2× Microswitches$3
PowerBattery pack (12V, 5A)$30-50
ChassisRobot platform kit$30-80
Total: ~$350-550 depending on choices

​
Software Dependencies

All software runs offline on the Raspberry Pi:
  • Ollama (local LLM inference): llama3.2:3b, moondream (vision)
  • Whisper.cpp (speech-to-text): ggml-base.bin model
  • Piper TTS (text-to-speech): en_US-lessac-medium voice
  • RPLidar SDK (obstacle detection): Python or Rust driver
  • ZeroClaw (AI agent runtime): Built from source

​
Safety Philosophy

The AI can REQUEST movement, but SafetyMonitor ALLOWS it. The safety system runs as an independent task and can override any AI decision:
  • Pre-movement checks: No obstacles in path?
  • Real-time monitoring: Slow down near walls
  • Collision response: Reverse if bump sensor triggered
  • Watchdog: Auto-stop if no commands for 30s
  • Hardware override: E-stop button cuts all motors
This prevents collisions even if the LLM hallucinates or loses context.

​
Use Cases

​
Play Companion

Child: "Let's play hide and seek!"
Robot:
  1. emote(expression="excited")
  2. speak(text="Okay! I'll count to 20")
  3. [waits 20 seconds]
  4. speak(text="Ready or not, here I come!")
  5. sense(action="scan")
  6. drive(action="forward", distance=1)
  7. look(action="find", prompt="a child hiding")
  ...

​
Patrol Mode

User: "Patrol the living room"
Robot:
  1. sense(action="scan", direction="all")
  2. drive(action="forward", distance=2)
  3. sense(action="motion")
  4. look(action="describe")
  5. [repeat]

​
Interactive Conversation

User: [speaks] "Hey Buddy, what do you see?"
Robot:
  1. listen(duration=5) → "Hey Buddy, what do you see?"
  2. look(action="describe")
  3. speak(text="I see a couch, a TV, and toys!")
  4. emote(expression="happy")

​
Personality: Buddy the Robot

The kit includes a default personality file (SOUL.md) that defines:
  • Playful: Enjoys games and jokes
  • Patient: Never frustrated
  • Safe: Always prioritizes safety
  • Curious: Loves exploring together
Example behaviors:
  • Counts to 20 for hide and seek
  • Stays 1 meter away unless invited closer
  • Alerts adults if child seems hurt
  • Remembers each child’s name and preferences
See Robot Kit Setup for customization.

​
Quick Start

​
1. Build Robot Kit

cd ~/zeroclaw
cargo build -p zeroclaw-robot-kit --release

​
2. Configure

mkdir -p ~/.zeroclaw
cp crates/robot-kit/robot.toml ~/.zeroclaw/
cp crates/robot-kit/SOUL.md ~/.zeroclaw/workspace/

# Edit for your hardware
nano ~/.zeroclaw/robot.toml

​
3. Install Dependencies

See Robot Kit Setup for detailed Ollama, Whisper, and Piper installation.

​
4. Test Components

# Test in mock mode (no hardware)
export RUST_LOG=info
./target/release/zeroclaw agent -m "Say hello and show a happy face"

# Test with real hardware (after wiring)
./target/release/zeroclaw agent -m "Move forward 1 meter"

​
Safety Notes

Before enabling real motors:
  1. Test in mock mode first
  2. Set conservative speed limits (max_speed = 0.3)
  3. Wire an E-stop button to GPIO 4
  4. Add bump sensors to GPIO 5, 6
  5. Test emergency stop works
  6. Supervise robot around children
  7. Enable LIDAR if available

​
Integration with ZeroClaw

The robot kit is currently a standalone crate. It defines its own Tool trait compatible with ZeroClaw but doesn’t require it. To integrate with core ZeroClaw runtime:
  1. Create adapter that maps robot-kit tools to src/tools::Tool
  2. Register in tool factory
  3. Load robot.toml config
  4. Enable safety feature for SafetyMonitor
See Robot Kit API for details.

​
Next Steps

Setup Guide

Complete hardware and software setup

API Reference

Tool specifications and examples

Raspberry Pi Setup

Configure Raspberry Pi GPIO

Supported Boards

Hardware platform comparison

​
Examples

Full example code is in crates/robot-kit/examples/:
  • basic_drive.rs: Move forward/backward/rotate
  • vision_test.rs: Capture and describe images
  • speech_test.rs: Listen and speak
  • patrol.rs: Autonomous room patrol
  • hide_and_seek.rs: Interactive game

​
License

MIT — Same as ZeroClaw

Build docs developers (and LLMs) love

Get started for freeTalk to us