Content Generation Pipeline

A common pattern: turn a topic or brief into finished, on-brand content through a chain of specialized agents — and scale it to many items at once.

The Approach

1. Researcher agent — enable the web_search and web_scraper capabilities so it can gather current, sourced material on the topic.
2. Writer agent — give it your brand voice. Encode the voice once as a skill and reuse it across every writer agent.
3. Editor agent — a separate pass for tightening, fact-checking, and formatting. Keeping write and edit as distinct agents produces noticeably better output than one mega-prompt.
4. Human-in-Loop — pause for a person to approve before anything goes live. The reviewer's comment is recorded on the execution. See Human-in-Loop.
5. Publish — on approval, an http_request node POSTs the content to your CMS, or an agent with the appropriate MCP server (e.g. Notion) writes it directly.

Scaling to Bulk Generation

To produce many pieces in one run, wrap the writer in a Loop node:

• Feed the loop a JSON array of topics (typed in, or produced by an upstream agent).
• Each iteration runs the writer on one item; results are collected into a single output.
• Use a faster/cheaper model inside the loop to keep cost and latency down.

Patterns Used

Pattern	Where
Specialized agents in sequence	researcher → writer → editor
Reusable brand voice	a skill shared across agents
Human oversight	Human-in-Loop before publishing
External delivery	http_request node or an MCP server
Bulk processing	Loop node