Switching from n8n to Claude Code: A Better AI Content Pipeline
Re‑implement the nine‑skill Claude Code pipeline, saving each step to a file, so drafts can read the original article, style guide, and research, reducing hallucinations and improving editorial consistency.
Deploy the Claude Code pipeline, replace n8n steps, and train writers to review side‑by‑side diffs for quick corrections.
Summary
Semrush’s content team moved from an n8n workflow to a Claude Code pipeline to address drafting failures caused by lack of full context. n8n could fetch research data but the drafting step produced hallucinations and inconsistent voice because it treated drafting as a single API call. Claude Code runs inside a folder, accessing the original article, style guide, research output, and past drafts simultaneously, and each skill writes an artifact file for easy re‑run. The nine‑skill pipeline—fetch article, research SERP, run EDI check, synthesize plan, identify outdated content, draft, compare, format—ensures drafts match Semrush voice, structure, and product details. Hallucinations are caught via side‑by‑side diffs, and the workflow now produces drafts that pass editorial review with minimal fixes, saving time and improving quality.
The article highlights that drafting needs persistent context and that workflow tools alone cannot provide the editorial reasoning required for high‑quality AI content.
Key changes
- n8n workflow failed drafting due to lack of full context
- Claude Code runs inside a folder, accessing all files simultaneously
- Pipeline has nine skills: fetch article, research SERP, run EDI check, synthesize plan, identify outdated content, draft, compare, format
- Each skill writes an artifact file for easy re‑run
- Hallucinations are caught via side‑by‑side diff
- Drafts match Semrush voice and structure
- Claude Code provides persistent context and reduces editing time
- Workflow now produces drafts that pass editorial review with minimal fixes