Snoze AI
An assistant that can
see the work.
It reads what you can read, and writes the answer back as something you can open.
01Rules
Four things it always does.
01
It reads what you can read.
The assistant runs with your permissions. A record you cannot open is a record it cannot see, so an answer never leaks something you were not allowed to know.
02
The answer is a saved view.
Ask what is at risk this week and you get a saved view in the workspace. Open it again next week and it is still right.
03
It cites the records.
Every claim links to the records it came from. When the answer is wrong, you can see which record misled it.
04
Violet means the machine did it.
Anything the assistant writes is marked in one colour, everywhere in the product. After a week you stop reading labels.
02Other models
And when it is not our model.
Your agents, this workspace
An MCP server exposes the workspace to tools you already run. They get your permissions and show up in the audit log.
Rules first, model second
Most of what people want from AI is a rule. Write the rule as an automation and keep the model for the part that needs judgement.
Bring your own key
Use the pooled allowance or point the assistant at your own provider.
Shape work. Then reshape it.
Free to start. No card. Describe your first workspace and it is ready in a minute.