n8n Official MCP vs Czlonkowski Unofficial
JeredBlugo watch the original →
the gist
Official n8n MCP adds 23 tools for AI agents to build/validate workflows remotely but rebuilds entirely on updates; unofficial offers partial edits and better executions tooling for efficient iteration.
Setup Process
The video demonstrates setting up n8n's official MCP server (public preview) by updating n8n, enabling MCP on specific workflows via the instance-level MCP button, and obtaining connection details. Users choose OAuth URL for Claude authentication or JSON with access token for IDEs/Claude Code. In Claude, add as remote custom connector: go to customize, plus button, paste OAuth URL, authenticate. This exposes 25 tools across Claude desktop/web/code/mobile.
Example prompt builds a cron workflow: "build me an NN workflow that runs every day at noon and checks my Gmail for email threads that I need to reply to If there are no threads that need to reply don't send anything If there are threads send me a telegram message with the full list." Agent uses get SDK reference tool then create workflow from code, generating TypeScript, validating/linting, converting to JSON, and adding to n8n canvas.
New Tools and Workflow Flow
Official MCP provides tools like get execution, get workflow details, validating workflow, publish workflow, test workflow, update workflow. Process: AI understands intent, writes TypeScript via code SDK, validates pre-runtime, converts to JSON for n8n, runs/tests. On failure, reads error and iterates (though author notes inconsistent success).
Recommend using Claude Code over desktop for better performance.
Official vs Unofficial Differences
Compared to Czlonkowski's unofficial n8n-mcp (GitHub: czlonkowski/n8n-mcp), official is remote (no Docker, works on phone), cleaner context (no token-bloating skills/docs). However, official update workflow rebuilds entire workflow from scratch, wasting tokens; unofficial has n8n update partial workflow for surgical edits.
Official get execution requires exact ID; unofficial executions tool lists/gets/deletes executions, easing debugging. Unofficial suits iteration/building; official better for initial creation.
Verdict and Use Cases
n8n excels for small, deterministic workflows (vs agentic platforms) to ration tokens/costs. Official MCP advances prompt-to-workflow but remains rough; pair with unofficial for scenarios. They complement: official remote, unofficial token-efficient iteration.