Quaddle: The 4-Servo Open-Source Robot Dog You Can Build in 20 Minutes — and Teach New Tricks by Hand

Petoi founder and OpenCat framework creator RZ Li shared their latest project on Reddit — Quaddle, an open-source quadruped robot dog kit driven by just four servos. The biggest selling point: no soldering required, no screws on the body (only four screws secure the legs), and the entire process from parts to a walking robot takes under 20 minutes.

This isn't just a moving toy — it's a complete robotics ecosystem built around the philosophy of "low cost, high learnability." For anyone wanting to get into AI robotics without being immediately discouraged by hardware complexity, Quaddle offers a genuinely thoughtful entry point.

A Walking Robot Dog in 20 Minutes

Quaddle's hardware design clearly prioritizes low barrier to entry. The whole machine runs on just four servos, weighs 170 grams, and measures 11 centimeters in body length — a true desktop-scale mini quadruped.

The no-solder, screw-free body design makes assembly feel more like snapping pieces together than building a robot. According to the creator, starting from raw components, going through assembly, calibration, and getting a gait running all takes under 20 minutes. For a DIY robotics project, that's an aggressive timeline — most comparable open-source projects take hours just for wiring and debugging.

Don't let the small size and low servo count fool you — its movement capabilities are surprisingly capable. Official specs put walking speed at roughly 1.8 body lengths per second, with sliding reaching 3 body lengths per second. It ships with over 90 built-in behaviors, including continuous rolling, three-legged walking, stair climbing, and even inverted walking. For a four-servo structure, achieving this range of motion says a lot about how much work went into the underlying gait algorithms.

A Clear Hardware Upgrade Path

Quaddle isn't a fixed configuration — it's designed with a clear progression: Builder (base model) → Buddy → Scout. The Scout version adds a fifth servo to drive a "radar"-style head structure, giving the robot dog environmental scanning capability.

On the sensor and compute side, the platform supports Arduino sensor expansion, with an optional Raspberry Pi Zero as a more powerful compute core, plus an optional AI vision module. This means users can incrementally upgrade based on budget and needs, evolving from a pure locomotion toy into a robot platform with perception and vision capabilities.

This layered design reflects the creator's core philosophy — making robotics accessible and affordable enough that people actually use it to learn. The four-to-five servo approach is a deliberate tradeoff between cost and functionality.

What is Raspberry Pi Zero? The Raspberry Pi Zero is an ultra-low-power single-board computer from the Raspberry Pi Foundation, measuring approximately 65×30mm and weighing just 9 grams. It features a single-core 1GHz processor and 512MB of RAM, typically priced around $15. In a robotics platform, it serves as an "edge compute core": compared to microcontrollers like Arduino, it can run a full Linux OS and support the Python ecosystem for more complex vision and networking tasks; compared to the more powerful Raspberry Pi 4 or Jetson series, it's far more size-, weight-, and power-friendly for a small quadruped. Paired with an AI vision module, the Pi Zero can run lightweight object detection inference to give the robot dog basic environmental awareness without significantly impacting overall weight or battery life.

Multiple Programming Paths and Hand-Guided Teaching

Quaddle's software openness is what truly sets it apart from ordinary toys. It offers a complete programming progression: graphical block-based coding (for beginners), Python, MicroPython, all the way to Arduino C++. This path covers virtually everyone from complete beginners to embedded developers.

The most interesting feature is hand-guided teaching. Using feedback servos, users can physically move the robot dog's joints to demonstrate a new motion — the robot records and reproduces it, no code required. This physical "record and replay" interaction dramatically lowers the barrier to creating new behaviors.

The creator also mentions leveraging AI agents to "vibecode" (generating code from natural language descriptions) new motions and iterate, bringing today's AI-assisted programming approach into the realm of robot motion design.

What is a Feedback Servo? A feedback servo is the core hardware behind the hand-guided teaching feature. Unlike standard servos that simply execute commands to reach a target angle, feedback servos contain built-in position sensors that continuously report the current joint angle back to the controller. This allows the robot to accurately record each joint's angle sequence even when passively moved by an external force, creating replayable motion data. This technology has long been used in industrial robotics as "drag teaching" — engineers physically guide a robotic arm through a trajectory to complete programming. Quaddle brings the same principle to a consumer-grade small quadruped, letting non-technical users create new motions in the most intuitive way possible, with no need to understand inverse kinematics or write any code.

Open-Source Strategy: Firmware, URDF, and 3D-Printed Parts

On the open-source commitment front, Quaddle follows OpenCat's established approach. Its firmware is based on the OpenCat framework and will be open-sourced; the robot's URDF model has already been published on HuggingFace to facilitate simulation and further development.

However, the creator is candid about one practical engineering tradeoff: some parts will be open-sourced with 3D-printable files, but not all components can withstand the mechanical stresses of locomotion when 3D printed — certain structural parts must be injection molded. For hobbyists hoping to build everything from scratch, this is a notable limitation. It remains a "semi-open" hardware product rather than a fully printable design.

What is URDF? URDF (Unified Robot Description Format) is the standard XML format in the ROS (Robot Operating System) ecosystem for describing a robot's physical structure, including link geometry, joint types, range of motion, mass, and inertia parameters. Publishing the URDF to HuggingFace means researchers can directly load Quaddle's digital twin into mainstream simulation environments like Gazebo, MuJoCo, or Isaac Sim — developing and validating gait algorithms and reinforcement learning policies without risking damage to physical hardware before transferring to real-world testing. This "simulation-first" development workflow has become standard practice in recent quadruped robotics research (as seen in ANYmal and Spot research papers). By opening up this infrastructure, Petoi significantly lowers the barrier for academic and hobbyist researchers to build on top of the platform.

The Tradeoffs of a Small-Scale Platform

At the end of the post, the creator poses two open questions: Is the four-to-five servo approach worth this particular accessibility tradeoff? And what directions would people most like to see explored further on a platform this size?

This touches on a long-standing tension in consumer robotics. Servo count directly determines degrees of freedom and motion richness, but also linearly drives up cost, weight, and power consumption. Quaddle's decision to cap at four servos trades motion precision and complexity ceiling for an extremely low barrier to entry and price. For education, beginners, and maker contexts, that tradeoff makes sense. For researchers pursuing realistic locomotion, a platform with more degrees of freedom may still be necessary.

OpenCat has been evolving since 2016, and Petoi has consistently pursued the mission of making robotics accessible and learnable for everyone. Quaddle can be seen as yet another aggressive push along that path — this time targeting cost and ease of use. It may not be the highest-performing quadruped robot out there, but it might just be the one most likely to get a newcomer genuinely up and running.