Build a Multi-Agent Research Pipeline with n8n (No Code Required)

Multi-agent AI workflows — where specialized agents handle different subtasks — are becoming the standard approach for complex automation. This guide shows you how to build one in n8n without writing code.

The pipeline we will build: a Research Agent that queries Perplexity for current information, a Writing Agent that drafts content from the research, and a Review Agent that checks factual accuracy and flags claims to verify.

Step 1 — Research Agent. Create an n8n workflow triggered by a manual form. Add an HTTP node calling the Perplexity API with your research topic. Parse the response to extract the key facts and sources.

Step 2 — Writing Agent. Pass the research output to a Claude AI Agent node. System prompt: "You are a content writer. Use only the provided research to write a [format] about [topic]. Include specific facts and cite sources."

Step 3 — Review Agent. Add a second Claude node with the system prompt: "Review this content for factual claims. List any claims that need verification and flag unsupported statements." Route flagged content to a human review Slack channel.

Step 4 — Connect and test. Wire the three agents sequentially using n8n's data passing. Add error handling nodes for API failures. Test with five topics before enabling for production use.