Multi-Agent Content Pipeline
Chain a researcher, a writer, and an editor agent so a single topic becomes publish-ready copy through specialized hand-offs.
What you'll build
Three specialized agents that hand off in sequence — a researcher gathers facts, a writer drafts from them, and an editor polishes and fact-checks. Splitting the work across focused agents beats asking one agent to do everything, because each has a narrow job and a tailored prompt.
Trigger: an API call or chat message with the topic in {{starter.input}}.
The workflow
Build it
Starter — accept the topic
Add a Starter. The topic arrives as {{starter.input}}.
Researcher agent — gather facts
Add an Agent and connect a Tavily block to it as an agent tool, so the agent decides what to search. Use Response Format:
{
"outline": ["string"],
"facts": [{ "claim": "string", "source": "string" }]
}Prompt: "Research the topic. Produce a factual outline and a list of claims, each with a source URL."
Writer agent — draft from the research
Add a second Agent. Feed it the researcher's output:
Topic: {{starter.input}}
Outline: {{agent1.output.outline}}
Facts: {{agent1.output.facts}}Prompt: "Write an engaging 600-word article following the outline. Use only the provided facts. Mark any claim you can't support with [needs source]."
Editor agent — polish and verify
Add a third Agent that receives {{agent2.content}} and the original facts. Prompt: "Copy-edit for clarity and tone. Verify every factual claim against the facts list. Remove or flag anything unsupported. Return clean Markdown."
Response — return the finished piece
Add a Response block returning {{agent3.content}}.
When you have several agents in one workflow, reference each by its block name — e.g. {{agent1.output}}, {{agent2.content}}. Rename blocks on the canvas (e.g. researcher, writer, editor) to make references self-documenting: {{researcher.output.facts}}.
Make it yours
- Add an image. Insert an Image Generator block after the writer to produce a hero image from the article title.
- Publish automatically. End with a Notion or WordPress-style HTTP block to push the final copy straight to your CMS.
- Quality gate. Add an Evaluator block to score the draft; loop back to the writer if the score is below a threshold.
- Different voices. Give each agent a different model — a strong reasoning model for the editor, a fast one for the writer.
Related
RAG Knowledge Agent
Answer questions strictly from your own documents using vector search, with a guardrail that blocks answers the sources don't support.
Lead Enrichment Agent
Turn a raw email address into a scored, enriched lead — pull company and contact data, have an agent qualify it, and write the result to your CRM.