AI Content Repurposing Agent
One long-form asset in. Five platform-native drafts and five title variants out, reviewed in Notion, announced in Slack.
I built an n8n agent that takes a creator's transcript, blog, or podcast notes plus their voice and audience, and in a single Claude Sonnet 4.5 call returns 5 platform-native drafts (LinkedIn, X, Newsletter, YouTube Short, Instagram) + 5 title variants, saved to Notion for review and pinged to Slack.
The workflow starts with a content intake form that creators can embed directly on a portfolio or landing page, they submit one long-form asset (transcript, blog body, or podcast notes) along with their voice, audience, and CTA. A Prepare Context step packages those inputs into a single prompt the agent can reason over.
The core is a single Claude Sonnet 4.5 call behind a strict JSON output parser: one request returns all five platform-native drafts and five title variants at once. Doing the whole job in one pass is reliable, fast, cheap, and trivially retryable, no orchestration of five separate model calls, no partial-failure states to clean up.
On the delivery side, a Flatten step unpacks the structured assets and each draft lands in a Notion review database (Title, Status, Content Type, Source URL, Audience, Summary, LinkedIn Post, X Thread, Newsletter, YouTube Short, Instagram Caption, Title Variants). A Slack notification then pings the review channel with the page link so whoever owns the voice can approve, tweak, or publish, no inbox hunting.
- n8n
- Anthropic (Claude Sonnet 4.5)
- Notion API
- Slack OAuth2
- Form Trigger