N8n MCP: AI agent toolkit for workflow generation, linting, and debugging

Built with n8n
1

Hi everyone - just launched n8n-mcp, a Model Context Protocol server that connects AI agents to your n8n workflows.

What it solves

  • Generate workflows from natural language prompts

  • Lint for subtle failures: deprecated nodes, schema regressions, silent data loss

  • Diagnose live executions with plain-language error explanations

  • Manage workflows programmatically from your IDE

Why it matters

If you use Claude, Cursor, or another MCP-compatible AI tool, you no longer need to switch to the n8n UI to build or debug workflows. Everything stays in your IDE.

How it works

Nine tools split into two groups:

Stateless (no n8n instance needed)

  • n8n_generate_workflow - natural language to workflow JSON

  • n8n_scaffold_node - build individual nodes with examples

  • n8n_lint_workflow - catch errors before you deploy

  • n8n_explain_execution - human-readable diagnosis of failed runs

Live-instance (requires n8n API key)

  • n8n_list_workflows, n8n_get_workflow, n8n_create_workflow

  • n8n_activate_workflow, n8n_list_executions

What makes it different

The linter and execution diagnosis focus on silent failures - unhandled edge cases, missing error branches, data loss in merge nodes. Not just “the node errored” but why, and what to fix.

Get started

Questions or feedback welcome below.