Connect AI to the real world
and let it learn
The tinyboxes are a playground for AI agents to interact with the physical and virtual world. Allow AI Agents to learn new skills in a sandbox environment, before deploying them in the real world.
Discover prebuilt
tinyboxes
We have a range of prebuilt tinyboxes that you can use to study the effects of different environments on AI and machine learning.
Wind
A tinybox that simulates the wind. It is used to study the effects of wind on the environment.
Optical
A tinybox that simulates the optical properties of the environment. It is used to study the effects of light on the environment.
Build your own tinybox
Build your own tinybox with Agent 2 Reality SDK. Connect AI to Reality with the tinybox AI Layer
from a2r import Sensor, Actuator, Agent # one import – runs on Pi or in sim
# 1 wire up reality in two lines
v = Sensor.voltage("cell_v", adc=0) # ADS1115 CH0 → volts
t = Sensor.temp ("cell_t", pin=4) # DS18B20 1-wire
fet = Actuator.pwm ("load_fet", pin=17) # N-MOSFET as electronic load
chg = Actuator.pwm ("charger", pin=22) # CC charger module enable
# 2️ ask the agent what you want to learn – plain English
lab = Agent(v, t, fet, chg)
lab.plan(
"Find the inflection where dV/dt spikes during a 0.1-1 A charge sweep "
"and return the current giving max Coulombic efficiency."
)
# 3️ press go – everything else is auto-logged
dataset = lab.experiment(duration="2h") # → Pandas DataFrame
dataset.save("battery_opt.csv").plot()Trusted by industry leaders
Our tinyboxes have been used by top companies for research and development in artificial intelligence and machine learning.
Let AI learn in safe sandbox
Create controlled environments where AI systems can safely interact with physical elements, testing responses and learning patterns without real-world consequences.
Accelerate Industrial research
Speed up your R&D cycle with purpose-built tinyboxes that simulate real-world conditions. Test AI applications in manufacturing, logistics, and quality control environments.
Create synthetic real data for AI training
Generate authentic, high-quality training data by capturing real-world interactions within controlled tinybox environments. Bridge the gap between simulated and actual data for more robust AI model development.
