Skip to content
Koidra
Book a demo

AI that acts on heat, mass and machinery

The agentic system for industrial processes.

One platform that sees everything your operation produces, decides what needs to happen, and, where the physics allows, runs it. Every step pays for itself before the next one starts.

Book a demo What autonomous operations actually takes

Proof

An AI ran a commercial crop better than expert growers. Twice.

Koidra won the international Autonomous Greenhouse Challenge twice, in 2018 and again in 2022. Each time the crop ran for months with nobody at the panel. That is the hardest version of the problem, and it is the reason we can use the word autonomous without an asterisk.

Aerial view of a commercial glasshouse complex, several ranges of glass with the packhouse and buffer tanks alongside
2×
Winner, Autonomous Greenhouse Challenge 2018 and 2022, a live crop each time
up to 27%
Yield improvement in commercial greenhouse operations
17%
Energy savings in commercial greenhouse operations
20%
Dryer efficiency gain at Ayo Biomass, after a two-phase KoPilot deployment

The platform

Four steps to autonomous operations. Start at any of them.

Most operations start by seeing themselves, or by moving maintenance off a whiteboard. Each step is a complete product with its own payback, and each one makes the next possible.

  1. Step 1 Essentials It sees itself
  2. Step 2 Maintenance Management It runs its own work
  3. Step 3 AI Advisory It explains itself
  4. Step 4 KoPilot It holds its own setpoints
  5. Step 5 Physical work and business decisions Not ours. Moving material is a robotics problem, and the commercial calls stay yours.

What autonomous operations actually takes, or see how the steps fit together.

One platform, because an agent can only act on what it sees

Koidra holds process sensors, energy meters, equipment history, maintenance work and the readings operators type on their phones in one place, against the same equipment and the same production batch. Most plants have this data. Almost none have it joined up, which is why almost none can automate anything above a single control loop.

Physics-aware AI, because a wrong setpoint is expensive

Our models are grounded in how heat, mass and machinery actually behave, rather than fitted to last quarter's numbers. The model knows a dryer cannot heat faster than its mass allows, so it never proposes it.

That is what makes an autonomous decision safe enough to trust with a live plant, and it is where our research heritage sits.

How it works

One reading, four things that happen next

A sensor value and a number an operator types are the same shape to the platform. That is why an alarm can fire on a hand-entered moisture reading, and why a trend can open a work order before anyone notices it.

How a reading becomes an action A sensor reading and an operator's hand-entered reading both enter one operating record. From that record an alarm reaches a phone, a work order opens against the right asset, reports are written, and KoPilot adjusts setpoints. Process sensor every few seconds Operator's reading typed on a phone One operating record same equipment, same production batch, same clock Alarm on a phone Work order opens Report is written KoPilot adjusts setpoints both are the same shape to the platform
A historian never sees the operator's reading, and a maintenance system never sees the trend. Holding both makes everything to the right of the record possible.

Industries

Built for operations where output and energy are decided hour by hour

Manufacturing

Inputs vary load to load and every machine downstream absorbs it. The drift surfaces on a utility bill a month later rather than as an alarm.

Watched here: Energy per tonne

Greenhouse Farming

Yield is what the year turns on, and it still depends on which grower is on shift. Light and heat follow the same setpoints nobody has revisited since last season.

Watched here: Yield per square metre, and energy per kilo

Data Centers

Cooling is the largest controllable cost in the building, and setpoints are still tuned by hand against a comfort margin nobody has revisited.

Watched here: PUE and approach temperature

See Koidra on your own plant data

Bring a week of readings from one line. We will show you what the platform sees, what it would have flagged, and what it would have scheduled.

Book a demo