Treating farms as intelligent systems capable of sensing, adapting, and regenerating.

AGROSTASIS™

One agentic
loop.

AGROSTASIS™ is the first agentic platform for agriculture that continuously senses, models, acts, and adapts to keep the field in balance as conditions change.

AGROSTASIS demo
The architecture

Perceives.
Believes.
Acts.

Sensing+

Soil / Plant assays · Geospatial · Imagery · Environment · In-field sensors

chemical · genomic · enzymatic

satellite · LiDAR · NDVI

drone · camera · autonomous microscope

weather · leaf/soil temp · RH

photonic · moisture · electrophysiology

Field State Model+
Running on simulated seasons

One model of the whole field. Soil chemistry, microbial community, enzyme activity, physical structure, water status and plant physiology are modelled as coupled processes on one daily clock, and every instrument is treated as a partial, intermittent observer of them.

Sampling Agent+
MVP · Lazy Dynamics

what to measure next to reduce uncertainty & obtain the highest-quality data

Input Agent+
In development

when & where to apply fertiliser, water, organisms

Automation Agent+
In development

plans & orchestrates the work on the ground

Active Inference

One symbiotic system.

An agent is any entity capable of modeling its world, pursuing goals, and solving problems.

Soil microorganisms, insects, and vines are.

We are building artificial agents with the same capacity.

They are built on Active Inference, a mathematical framework that gives us the ability to model how agents act to resolve their own uncertainty.

Our agents run on the edge: in the field, on local hardware, alongside the biology they work with.