The platform
One intelligence layer above the hardware already in the field.
Terra connects existing cameras, sensors, weather sources and farm systems, reconciles what they produce into one record, reasons over it, and reports what is worth someone's attention, with the evidence attached.
End to end
Six stages, and only the first one is yours to buy.
The farm supplies the equipment and the ground truth. Terra supplies everything between the raw feed and the decision.
Yours
Existing farm hardware
Whatever is already in the field. Terra begins here rather than with a purchase order.
- Cameras
- Sensors
- Weather stations
- Drones
- Gateways
- Farm APIs
Ingestion
Terra Connect
Devices, streams and data sources are connected over the protocols they already speak, with credentials and access scoped per source.
- MQTT
- RTSP / ONVIF
- REST & webhooks
- Imports
Ingestion
Data normalisation
Units, sampling intervals, timestamps and identities are reconciled, and every reading is bound to the zone and crop it belongs to.
- Unit alignment
- Time alignment
- Zone binding
- Quality flags
Reasoning
Terra Intelligence
Computer vision, sensor fusion, analytics and prediction run over the combined record rather than over one feed at a time.
- Computer vision
- Sensor fusion
- Analytics
- Prediction
Model
Farm digital twin
The farm as a structured, continuously updated model: fields, zones, crops, instruments, conditions, history and open risk.
- Fields & zones
- Crops
- Instruments
- State & history
Output
Risks, insights, recommendations
Ranked and explained. Every output names the signals behind it, states its confidence, and says what it recommends confirming.
- Risk scores
- Anomalies
- Predictions
- Priorities
Terra Connect
Most of the difficulty is in the joining.
Instrumentation arrives one supplier at a time, and each brings its own units, intervals, identifiers and portal. Reconciling them is unglamorous and it is the precondition for everything downstream.
Units and intervals reconciled
Two probes from two suppliers report the same quantity in different units, at different intervals, with different notions of a timestamp. Until that is resolved, a comparison between them is arithmetic on incompatible numbers.
Every reading bound to a zone
A measurement without a place is a number. Each source is registered against the zone it observes, so a reading, a detection and a forecast can be brought together because they describe the same ground.
Data quality separated from agronomy
Flatlined channels, implausible steps, drifting calibration and silent devices are flagged as instrument problems. A failing probe should never leave the platform as a soil condition.
History retained
Zones are judged against their own record rather than a general expectation, which is what makes a slow drift visible long before it crosses any fixed threshold.
Terra is designed to integrate with compatible hardware, protocols and data sources. Compatibility is not universal and is not assumed: which of your devices can be connected, and what each of them can contribute, is established during the farm assessment before anything is committed.
Terra Intelligence
Signals that mean little apart, read together.
The reasoning layer correlates sensor readings, camera detections, weather, geography and history on the same zones and the same timeline. What comes out is a risk with its evidence, not another feed to watch.
Signals
- Low soil moisture
- High temperature
- Canopy stress in imagery
Elevated water-stress probability
Raised as a probability with the contributing readings attached, and routed to irrigation priorities.
Signals
- High humidity duration
- Leaf symptoms detected
- Favourable temperature range
Elevated disease risk
Reported as a suspected condition requiring agronomic confirmation, never as a diagnosis.
Signals
- Rising pest counts
- Adjacent zones affected
- Conditions favour development
Possible infestation spread
Direction and pace are estimated from the zones involved, and inspection is prioritised accordingly.
Terra Twin
A model of the farm, kept current.
Fields, zones, crops, cameras and sensors are held with their relationships intact, each carrying current state, measurement history, detected risks and predicted conditions. A zone with no instrumentation appears as structure without measurement rather than as an estimate.
Model structure
Example Farm
2 fields · 5 zones · 56.7 ha
Field North
2 zones · tomato
Zone A
12.4 ha · 3 cameras · 6 sensors · healthy
Zone B
15.1 ha · 2 cameras · 5 sensors · water stress
Field South
3 zones · pepper, cucumber
Zone C
9.7 ha · 2 cameras · 4 sensors · healthy
Zone D
11.2 ha · 4 cameras · 4 sensors · pest risk
Zone E
8.3 ha · 2 cameras · 3 sensors · monitoring
Terra Optimize
The output is a working list, not a wall of telemetry.
Scored risks become an ordered set of recommendations: which zones to walk, where water is most needed, which blocks warrant closer monitoring. Each one states the reasoning behind it and what still needs confirming.
Terra does not operate irrigation systems, dose chemicals or control machinery. Write-back integrations with farm equipment are a future capability rather than part of the platform today.
Recommended next actions
Irrigation review
Zone B · priority High
Moisture below the zone's own baseline for four days, with canopy stress rising in imagery and three days above the usual maximum.
Scouting priority
Zone D · priority High
Detection counts up across three capture cycles, with the neighbouring zone showing the same direction of travel.
Instrument check
Zone E · priority Medium
Moisture channel static for 41 hours while neighbouring probes responded to the same rainfall. Treated as a data-quality problem, not a soil condition.
How Terra states what it knows
Agriculture is biological. The platform speaks accordingly.
There is no accuracy figure on this site, no guaranteed yield and no promise of prevention, because none of those could be published honestly. What can be published is how findings are expressed.
Probability, not verdict
Risk is reported as a probability with the signals that produced it. A platform that reports biology in certainties is making a claim it cannot support.
Confidence, stated
Every detection and every risk carries a confidence value, and low confidence is shown rather than filtered out of view.
Suspected, pending confirmation
Visual findings are suspected conditions requiring agronomic confirmation. Terra directs an inspection; it does not conclude one.
Evidence attached
Each alert lists what was observed, the frames or readings behind it, and what it recommends confirming, so a finding can be checked rather than trusted.
Deployment principles
How a Terra deployment is put together.
No hardware lock-in
Terra is designed to read equipment a farm chose for its own reasons, and a farm that later replaces a supplier keeps its history and its zone model.
Credentials scoped per source
Each connected camera, gateway or API is registered with its own access, so a source can be added or revoked without touching the rest of the estate.
Farm data belongs to the farm
Deployment defines where a farm's data lives and who reaches it. Terra exists to interpret an operation's own record, not to accumulate one.
Built to be checked
Findings carry their evidence, instruments carry their health, and the platform states which source a figure came from, including when that source is a third-party forecast.
Next step
Request a Farm Assessment
Every agricultural operation has a different data footprint. Terra deployments are scoped according to the farm, existing hardware and required intelligence capabilities.
The assessment establishes what your existing hardware can support before anything is committed.