Skip to content
Ghost AI WriterEditorial control plane Operator sign in

Implemented capabilities

Every handoff stays visible.

The product connects reviewed scope, content generation, Ghost outcomes, usage evidence, recovery and static delivery without pretending those steps are the same thing.

Prepare

Review exact run scope

Use manual keywords or explicit pending topic IDs, deduplicate them, inspect the immutable manifest and keep every batch within the validated 25-item limit.

Generate

Create non-commodity drafts

Intent, direct answer, article body, metadata, originality and humanization stages keep their recorded backend identity and metered usage.

Publish

Keep draft and live distinct

Ghost draft is the safe default. Live publication requires per-run acknowledgement, while uncertain writes reconcile before another POST.

Observe

Watch real pipeline events

Run monitoring uses backend-emitted stages, keyword outcomes, token usage, known cost and linker results instead of decorative timers.

Recover

Resume without rewriting history

Cancellation preserves completed work, retries preserve the original manifest, duplicate-index fallback remains visible, and portable backups are verified before download.

Deliver

Build one verified artifact

Published Ghost posts become sanitized static output, committed to the GitHub mirror and delivered canonically through Cloudflare.

Product boundaries

Control without false promises.

  • No public signup or subscription billing exists.
  • Ghost remains the content editor and publishing authority.
  • CMS synchronization is read-only and operator-triggered.
  • Soft budgets warn but never stop a run.
  • Railway and Ghost hosting costs are excluded from AI usage.

See the operating model

Follow each control from setup to delivery.

Open documentation