Describe the internal tool in a conversation. An agent builds it using any data in the workspace and publishes it privately. There is no code repository, no deployment and no ticket to raise.
Write one sentence in a stream saying what is needed. The agent finds the data anywhere in the workspace, builds the screens and shows the result. Reply with changes and it updates the app.
Mini apps are organised in folders, each with an owner, a last-modified date and a history. They follow the same permissions as everything else, so an app can never show someone more than they are already allowed to see.
No code repository, no build step and no waiting in someone else's queue.
The app reads and updates the same data people work in.
Organised like files, with a history and an owner.
Install an app that already exists instead of describing it again.
A mini app can only see what its user and its agent are allowed to see.
Shared inside the workspace, never on the public internet.
Ask for a change and the agent looks at the app, the data behind it and how people use it, then rebuilds it and shows what changed.
No. Describe what is needed in a sentence. A developer can extend the app later through the API, but building one needs no code.
Yes, if it has permission. The result is a draft that a person publishes.
The app follows the data. It reads the source directly, not a copy, so a new field is available immediately and a removed one is flagged.
Write one sentence in a stream and an agent has it ready that afternoon.