How-To · · 5 min read
Build a Brand Kit from a public website
The onboarding worker extracts structured identity, voice, audience, and visual signals from public pages.
Last reviewed:
Website analysis runs as a bounded background job. Public-address checks block local and private network targets before the worker fetches anything.
When extraction completes, the structured result is applied to the default brand and becomes available to writing, social, image, and chat context. Failed runs stay recoverable and never replace the current Brand Kit with partial output.
Secondary brand records can be managed from Brand Kit. When separate products require a hard data and generation boundary, use separate tenant-isolated workspaces rather than relying on an in-workspace selector.
Frequently asked questions
What does website ingestion extract?
Structured identity, voice, audience, and visual signals from public pages. The result becomes generation context for writing, social drafts, images, and the guided chat roles.
Is the ingestion safe to point at any URL?
Ingestion targets public websites only: public-address checks block local and private network targets before the worker fetches anything, and the whole run is a bounded background job.
What happens if extraction fails midway?
Failed runs stay recoverable and never replace your current Brand Kit with partial output — the kit you had remains intact until a successful extraction completes and is applied.
Can one workspace hold multiple brands?
Secondary brand records can be managed in a workspace, but supported generation flows use the workspace's active/default context rather than guaranteeing per-request selection. Use separate tenant-isolated workspaces when identities require a hard boundary.