Permute connects business data so you can query it, publish reusable transformations, build apps, and run approved actions. These workflows share resource permissions; each app, user, and API key still needs the access required for its work.
Organizations and workspaces
An organization contains your team and workspaces. Workspaces contain connected data and the resources built around it. Choose a workspace before making data requests; selecting one does not grant access to its contents. See Permissions and sharing for access rules.
Sources and stored data
| Concept | Use it for |
|---|---|
| Connector | Access to an external system. Available tables and sync or live-query behavior depend on the provider. |
| Dataset | An imported file or related application tables supplied through the API or SDK. |
| Managed database | Application-owned records with a published schema and version-checked writes. |
| Source | A common way to discover and inspect data available for reads, including connectors, datasets, databases, and operation outputs. |
Start by discovering sources and inspecting their tables. Use returned IDs and exact table and column names. A connected source can still be preparing its first data or serving an older result after a failed refresh; inspect availability before relying on it.
Queries, evidence, and answers
Use structured queries for exact rows, counts, and aggregates. Use evidence search to find supporting material and citations. Answer jobs interpret business questions through the server SDK or REST API; compatible AI clients can reason over permitted data through MCP.
Data Flows
A Data Flow connects transformation steps into a workflow that applies declared logic to inputs and produces reusable outputs. Published outputs can become sources for later queries and apps (a saved draft alone does not refresh the data).
Keep business definitions in facts and shared skills where the workflow supports them, and encode repeatable calculations in tested transformations. Definitions and instructions describe intended behavior; verify the resulting data before publishing.
Apps and Business Actions
A hosted app serves a frontend you build in your own project. Ordinary app reads require access for both the viewer and app. Use a managed database for app-owned records and a published Business Action when you need approved business logic or bounded writes. For an analytical UI, follow the dashboard app example to display governed metrics and customer breakdowns.
A Business Action declares its inputs, outputs, and allowed work. Authorized callers can run that contract without receiving direct access to its underlying sources. Start with the app guide or Business Actions guide for their publication and access requirements.
Drafts, runs, and publication
A draft stores proposed changes. A run executes logic against data. Publication selects the definition used by live callers or scheduled work. These are separate steps: saving does not execute, and publishing an operation does not by itself prove that its data is current.
Restoring a revision recovers a definition; it does not undo record changes or external effects. Follow each workflow’s test and publication steps. Write Sync has its own approval contract: publishing a reviewed preview starts delivery and enables future captures.
Background work has two execution modes: caller access follows the initiating caller’s permissions; automation access uses authority approved for the automation and can continue after its configurator signs out. See background-work permissions for execution identity, approved access, and control of scheduled work.