CloudKites
CloudKites
AI-native imaging, where care happens.
Trustworthy clinical AI that runs on-site — beginning with AI-native medical imaging, growing into an agentic, and ultimately embodied, clinical assistant.
Designed for clinical trust
Four principles hold across everything we build.
Why on-site
Care happens in the room — so the AI lives there too
We made a deliberate choice early on: the intelligence belongs where the work is, not somewhere far away that a clinician has to trust on faith. That single decision shapes almost everything else about how CloudKites is built.
When analysis runs on-site, the most sensitive thing in the building — patient imaging — never has to leave it. There is no round trip to somewhere else, no copy of a scan resting on hardware you do not control, no dependence on a link staying up at the exact moment a clinician needs an answer. The facility keeps custody of its own data, by design and by default, and that custody is something you can point to rather than something you are asked to believe.
On-site also changes the felt experience of the tool. Assistance arrives at the pace of the procedure, not the pace of a network, which is what lets it sit quietly inside the moment instead of interrupting it. A clinician working through an endoscopy, reading a slide, or studying a scan should never be waiting on the software — the software should be waiting on them. Keeping the work local is what makes that rhythm possible.
And because the system is local, it is legible. Every finding it surfaces can be reproduced, logged and reviewed inside the same walls where it was made — the same input giving the same structured output, every time, available to the people who are accountable for the read. On-site is not a feature bolted on at the end; it is the foundation the rest of the platform stands on.
One platform across endoscopy, pathology and radiology.
Real-time assistance during the work that matters — on the equipment you already use — with structured findings flowing to the record. The clinician makes every call.
- EndoPod — real-time AI for endoscopy.
- PathoPod — assisted digital pathology.
- TomoPod — AI-native radiology imaging.
How a read works
Four steps, the clinician in command of each one
From the moment an image is captured to the moment a finding reaches the record, the path is the same across every modality — and a human is in charge at every point along it.
It begins with capture, on the equipment a department already uses. Nothing about the way a clinician works has to change for the assistance to be useful; the platform meets the existing instrument where it is rather than asking a team to adopt a new way of doing the job. The work that matters stays exactly where it was.
Then comes the assistive step — the part people imagine when they hear the word AI, and the part we are most careful about. The system reads what is in front of it and offers what it sees: a region worth a second look, a structure worth measuring, a pattern worth naming. It offers; it does not decide. Every one of those proposals is held lightly, ready to be confirmed, refined or set aside by the person reviewing it.
Review is where judgement lives. The clinician brings context the image cannot hold — the patient, the history, the reason for the study — and weighs the proposal against it. Only after that human review does anything reach the record, as a structured, reproducible result that the next clinician can read and trust. Four steps, one principle running through all of them: the software assists, and the clinician decides.
Free for everyone
Genuinely useful clinical AI you can download today.
The family of products
One idea, expressed across every place imaging is read
CloudKites is not a single app. It is a family — three imaging products on a shared foundation, an assistant that works across them, and a set of free tools that put genuinely useful clinical AI in anyone's hands.
At the centre sit three imaging products — EndoPod for endoscopy, PathoPod for pathology, TomoPod for radiology. They look different on the surface because the work they support is different, but underneath they share one AI-native foundation and one set of guarantees. Whichever modality a department starts with, it inherits the same on-site analysis, the same assistive posture, and the same auditable record. The trust model does not get diluted as you add modalities; it stays whole.
Around those products is Myro, an assistant that listens, sees and reasons alongside the clinician — handling the busywork that surrounds a read and surfacing what matters, while always leaving the decision to a person. And beyond the clinical platform sits a deliberately free layer: MedPodGP for general practice and Emu for everyone, on-device tools we give away because useful clinical AI should not be gated behind a budget. Together they are one idea, expressed wherever imaging is read and care is delivered.
Where we're heading
From AI-native imaging to an agentic clinical assistant, and ultimately to embodied, physical AI.
Bring clinical AI on-site.
Talk to us about endoscopy, pathology and radiology — or download the free tools today.