Agentic Content Pipeline - Research to Published, Gated

An agentic workflow is a system where the AI decides the next step itself, repeatedly, until the job is done — not a single prompt, and not a fixed Zapier chain.

My own SEO engine, live on Railway: a daily pipeline that turns raw sources into published articles, where a human only ever presses one button.

A scheduled intelligence step reads the sources and produces ranked topic suggestions.

A topic-planner agent groups them into clusters and picks what to write next.

A writer agent (Claude) drafts the article against the brief.

A deterministic quality and fact-checking step then runs with an auto-fix loop, and it blocks fabricated statistics. This is code, not a prompt asking the model to be careful — that is the difference between a demo and something you can leave running.

Internal links are added automatically, and a hero image is generated.

The flow is draft, then approve, then image, then publish. Publish is the only step that touches the live site.

The whole thing has an admin review dashboard.

Next.js, Drizzle, Postgres, Claude API, scheduled jobs, and the GitHub API to Vercel for publishing.

It publishes to a static site, not to WordPress. That connector is the piece I would add for a client on WordPress.

This is my own product, not client work. The architecture — autonomous steps plus a hard human gate before anything goes live — is what I would build for your workflow.