Morning brief
Assembled and waiting for you to open it
Technology
How Ora works.
Ora is an AI agent computer built around a cloud AI architecture. The device sits on your desk and runs the Ora experience. The models run in the cloud, through a provider account you connect with your own key.
This page describes architecture and design intent. It does not publish a figure that has not been measured, and it does not name a provider Ora has not yet connected to.
How it connects
Three parts, one system.
Ora is the machine and the software. The thinking happens at the AI provider you choose, on your own account. Keeping those three apart is the whole design.
- 01On your desk
Ora device
The machine you own. It stays on, holds your files and keys, and runs the Ora experience.
- 02The software
Ora experience
Where you set up agents, review their work and schedule automations. It runs on the device.
- 03The thinking
Cloud AI provider
The AI model your agents call, using your own provider account. Ora does not resell AI usage.
Your work stays on your device. When an agent needs a model, Ora sends that request to the provider you connected and brings the answer back. Ora does not run large AI models on the device itself.
- On your device
The experience, your files, your key
The Ora experience runs on the machine in front of you. Your documents, your agent setup and the provider key you connect stay on the device.
- In the cloud
The model that does the thinking
When an agent needs a model, that request goes out to the AI provider you connected, on your own account. The provider bills you directly.
- Not on the device
Large model execution
Ora is not a model and does not run large AI models locally. It is the machine, the software and the connection — the intelligence is called, not installed.
Capabilities
What you actually do with it.
Four ideas carry the whole experience. None of them require you to open a terminal or edit a configuration file.
Agents
An agent is a job you describe once — a reader, a researcher, a drafter. You give it instructions and the files it is allowed to touch, and it keeps working the same way every time you call on it.
Each agent keeps its own instructions and its own history, so you can see what it actually did rather than guess.
Automations
An automation is an agent on a schedule or a trigger. Because the device stays on, work can run at six in the morning or every time a file lands, with nobody sitting in front of it.
This is the part a browser tab cannot do: when the tab closes, the schedule goes with it.
Workspace
A workspace is where your projects, source files and the outputs your agents produce live together on the device. Results arrive next to the material they came from.
Files stay local. Nothing is synced to Ora, because Ora does not operate a storage service.
Provider connection
You bring an account with a cloud AI provider and add its key during setup. Ora uses that account to make the calls your agents ask for.
Ora does not resell AI usage and takes no margin on it. Your provider's bill is between you and them.
The software
The experience, screen by screen.
Agents, automations, the workspace and the provider connection, as they are being designed. Every screen here is a concept rather than shipping software, and is labelled as one.
Ora Home
Provider connectedOra Home
Your intelligence, agents and recent work in one view.
Ora intelligence
Ready
Active agents
1 runningInbox triageRunning
Weekly reportScheduled
Research notesIdle
Recent work
Summarised this morning's unread threads08:40
Drafted a supplier follow-up for review08:12
Filed new documents into ProjectsYesterday
Ask Ora to do something…
Summarise my morningDraft a replyFile this to Projects
Next scheduled run
Weekly report, Friday evening
On this device
Your keys and your files stay local
Your intelligence, agents and recent work in one view.
Design concept — not shipping software
Design intent
Built to be left running.
A machine that holds your agents has a different job from a laptop. These are the decisions that follow from that, and beside each one the figure that would prove it.
Always on
Ora is designed to run continuously rather than in bursts. An agent that only works while you watch it is not much of an agent.
Sustained-load figuresNot publishedQuiet by design
Thermal design comes before peak numbers. A machine that lives on a desk and never sleeps has to be tolerable at two in the morning.
Measured acousticsNot publishedOne designed object
The intent is that the enclosure, the thermal path and the board are designed together, rather than a case bought around a board.
Enclosure materialsNot publishedTwo configurations, one architecture
Ora Core is the entry configuration and Ora Mini is the higher one, with twice the memory and storage. The architecture does not change with the size of the box.
Per-device thermal figuresNot published
Still to be announced
The gaps are deliberate.
Everything below is unconfirmed. A plausible number or a familiar provider logo here would be easier to read and impossible to trust, so neither is here.
- Supported AI providersNot announced
- No integration with any provider is operational yet. Ora will name providers once a connection has been built and tested, not before.
- Release version and dateNot announced
- The Ora experience is in design. The screens on this page are concepts, and the shipping build has not been dated.
- Model and token costsNot announced
- What you pay for AI usage is set by the provider account you connect. Ora does not set those prices and cannot quote them.
- BenchmarksNot announced
- Ora will publish performance results when they come from a repeatable test on final hardware, with the method printed beside the number.
Open hardware figures
- Networking
- To be confirmed
- Ports
- To be confirmed
- Dimensions
- To be confirmed
- Weight
- To be confirmed
- Power
- To be confirmed
- Acoustics
- To be confirmed
- Warranty
- To be confirmed
Target configuration, pending hardware compatibility and BOM approval. Not a final certified specification. If something you need is still blank, tell us — that is how the blanks get prioritised.
Next
Two devices. One architecture.
- Ora Core
- For a first dedicated AI computer. Everyday agents, documents and automations on one quiet machine.
- Ora Mini
- For larger workspaces and longer-running automations. More memory and storage for keeping more work and more agents open at once.