n8n-local-setup

star 1

Guide AI agents through safe local n8n setup with Docker Compose, Postgres, Compose ngrok, hosted n8n on Hostinger Coolify VPS, launcher/menu use, skills-first routing, and optional AI-coding-agent MCP feature references. For any n8n workflow, helper-script, or live n8n task, first apply n8n-agent-rules.

weijunswj By weijunswj schedule Updated 6/14/2026
play_arrow Run Skill in Manus View GitHub

name: n8n-local-setup description: Use when setting up safe n8n environments: localhost/ngrok dev stack, production Cloudflare Tunnel self-hosting for local/CGNAT machines, hosted n8n on Hostinger Coolify VPS, launcher/menu use, official n8n Skills setup, or instance-level MCP references. For workflow, helper-script, or live n8n tasks, first apply n8n-agent-rules.

n8n Local Setup

Use this skill when the user needs to set up or explain local n8n, the Docker Compose local dev stack, Compose ngrok dev tunneling, the separate production Cloudflare Tunnel self-hosting stack for local/CGNAT machines, hosted n8n on Hostinger Coolify VPS, the guided launchers/menus, official n8n Skills routing, or official instance-level MCP references.

For Hostinger VPS plus Coolify setup or server maintenance, use codex-ssh-hostinger-coolify-setup-maintainer first. Return to this skill for n8n-specific deployment guidance after Coolify exists.

For any n8n workflow, helper-script, official n8n Skills, their entry-point meta-skill use, n8n_live, import/export, validation, credential, webhook ID, activation, execution, repo/live sync, or live n8n task, apply n8n-agent-rules first.

Source And Runtime Material

Core Rules

  • Use the local full-fidelity references in references/n8n/ for local dev setup, updates, Compose ngrok dev tunneling, production Cloudflare Tunnel self-hosting for local/CGNAT machines, Docker Compose plus Postgres, and hosted n8n Hostinger Coolify VPS details.
  • Treat the Compose ngrok service as the supported dev tunnel path for local webhook testing.
  • Treat the production Cloudflare Tunnel stack as the production self-hosting path for local/CGNAT machines. Keep it separate from the local dev stack and menu.
  • Do not treat TryCloudflare or Quick Tunnel as production hosting. If a quick tunnel smoke test is ever useful, keep it as a separate future smoke-test-only follow-up.
  • Use the local stack templates in templates/local-stack/ for local n8n + postgres.
  • Treat templates/local-stack/_n8n-local.cmd as the recommended local stack entrypoint for guided start, update checks, logs, status, URLs, Postgres backup packages, and local backup restore actions.
  • Use templates/production-cloudflare-stack/ for production self-hosting from a local/CGNAT machine through Cloudflare Tunnel.
  • Treat templates/production-cloudflare-stack/_n8n-production-cloudflare.cmd as the production Cloudflare stack entrypoint for preflight, start/stop, restart, status, logs, Postgres backup, image updates, and production URL display.
  • Use references/ai-agent-platforms/ for Codex, Claude Code, OpenCode, Antigravity, ChatGPT web, Claude web routing, official n8n Skills installation notes, and official instance-level MCP references.
  • Use the official n8n-io/skills plugin instructions for Codex and Claude Code where plugin hooks are supported. Marketplace registration alone is not installation: before reporting setup complete, verify n8n-skills@n8n-io is installed and enabled in the host plugin list, not merely available. On Windows, immediately run node repo/scripts/repair-codex-plugin-windows-hooks.cjs --plugin-root "<installed-plugin-cache-path>" --windows --write --plugin-id n8n-skills@n8n-io, then node repo/scripts/audit-n8n-skills-plugin-hooks.cjs --plugin-root "<installed-plugin-cache-path>" --windows --verify-output, before approving or trusting hooks. Do not treat a temporary marketplace checkout such as .tmp\marketplaces\n8n-io\plugins\n8n-skills as the installed plugin cache.
  • The Windows hook repair wrapper invokes explicit Git Bash only: C:\Program Files\Git\bin\bash.exe or C:\Program Files\Git\usr\bin\bash.exe. It rejects WSL/System32 Bash, including C:\WINDOWS\system32\bash.exe.
  • If Codex or Claude Code opens session-start.sh on every new chat, the installed plugin still has an unsafe bare .sh hook on Windows. Run the repair command above against the installed plugin cache. If repair fails, audit fails, or hook JSON output verification fails, do not approve those hooks; use npx skills add n8n-io/skills and keep the target repo cue that loads using-n8n-skills before n8n work until the plugin can be repaired or updated.
  • For Antigravity/AG2, install or copy the official upstream n8n skill folders into the Antigravity plugin-scoped skill path, such as C:\Users\<user>\.gemini\config\plugins\n8n-skills\skills\<skill-name>\SKILL.md; verify using-n8n-skills\SKILL.md exists there before relying on Antigravity to load it.
  • Plain skill installs do not include the plugin SessionStart, PreToolUse, or PostToolUse hooks. Local Antigravity plugin-scoped folder installs also do not include the official n8n plugin hooks. Make sure the target repo AGENTS.md cues the agent to load the official n8n Skills entry-point meta-skill, currently using-n8n-skills, before n8n work.
  • Official n8n Skills plus official instance-level MCP setup/config references are secondary and not part of the beginner local setup path.
  • Use skills/ai-coding-agent-rules for generic AI coding agent rules.
  • Use skills/n8n-agent-rules or references/n8n-agent-rules.md for the full n8n operating ruleset.
  • Keep tokens, API keys, webhook secrets, tunnel tokens, .env values, real domains, account IDs, DNS values, IPs, backups, exports, certs, and private deployment notes out of repo files and chats.
  • Do not expose Postgres publicly.
  • Do not expose n8n directly on public host port 5678.
  • Do not run live n8n import/export, workflow activation, execution, publish, unpublish, archive, delete, or credential actions from this toolkit repo.
  • Do not run live Cloudflare, DNS, tunnel, Docker, credential, workflow activation, import/export, or production actions without explicit current-turn approval naming the target and allowed operation.
  • For live n8n or production work, require explicit current-turn confirmation and identify the target instance first.

Common Outputs

Workflow

  1. Confirm the user's target platform and whether they want docs, templates, or a local setup diagnosis.
  2. Read the relevant local reference file inside this skill folder when exact setup detail matters.
  3. Point the user to the AI-facing template, skill, or reference when they need copy-ready material.
  4. Keep live actions safety-gated and separate from repo/docs/template work.
Install via CLI
npx skills add https://github.com/weijunswj/ai-agent-toolkit --skill n8n-local-setup
Repository Details
star Stars 1
call_split Forks 1
navigation Branch main
article Path SKILL.md
More from Creator
weijunswj
weijunswj Explore all skills →