How Promptic is organized

Promptic uses a hierarchy so teams can share administration without mixing application data.

Organization

An Organization is the administrative home for a team. It owns:

• organization members and administrative roles;
• the billing account, plan, invoices, and aggregate usage;
• organization-wide provider credentials and model availability;
• one or more AI Applications.

Organization access does not automatically mean every member should receive every AI Application credential. Use AI Application membership and keys as the narrower operational boundary.

AI Application

An AI Application represents one application, agent, workflow, or production use case. It is the billable capacity unit and contains:

• AI Components;
• traces and spans;
• AI Application members;
• API keys;
• model-provider overrides;
• usage attributable to that application.

AI Component

An AI Component is one measurable AI-powered capability inside an AI Application. Examples include an email classifier, support agent, extraction prompt, or tool router.

The component is the scope for most operational data:

Tool-optimization components use the experiment lifecycle but produce tool descriptions rather than a deployable prompt.

Every trace belongs to the AI Application authenticated by its ingestion API key. OpenTelemetry service.name and deployment.environment.name identify operational service and environment dimensions. Spans can optionally be attributed to AI Components independently of optimization, and one trace may contain spans from several components or services.

IDs and names

Names help people navigate the dashboard. Stable IDs are the integration boundary used by the SDK and API. Renaming a component does not change its ID.

Use names that describe business function, not temporary implementation detail. Prefer invoice-extractor over gpt-4-test-3.

Where configuration applies

Next, read AI Components or create the structure with the Quickstart.