n8n-MCP
Gives AI assistants deep knowledge of 2,000+ n8n nodes, plus optional workflow management on your own n8n instance.
Add to your client
Copy the config for your MCP client and paste it into its config file.
npx n8n-mcpPaste into ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"n8n-mcp": {
"command": "npx",
"args": [
"n8n-mcp"
],
"env": {
"MCP_MODE": "stdio",
"LOG_LEVEL": "error",
"DISABLE_CONSOLE_OUTPUT": "true",
"N8N_API_URL": "<your-n8n-instance-url>",
"N8N_API_KEY": "<your-n8n-api-key>"
}
}
}
}Step-by-step guides: Add to Claude Desktop · Add to Cursor · Add to Windsurf
Before you start
- Node.js for npx (or Docker for the container image)
- Optional: a self-hosted or cloud n8n instance for the management tools
- Optional: n8n API key from Settings → API in your n8n instance
About n8n-MCP
n8n-MCP ships a pre-built SQLite database describing every n8n node — properties, operations, docs, and AI-tool variants — so search_nodes and get_node answer instantly and offline. Detail levels are token-aware: minimal (~200 tokens) for metadata, standard for essential properties, full (~3–8k tokens) when the agent really needs everything.
The intended build loop is templates-first: search 2,352 community templates by task, node types, or metadata (complexity, setup time, target audience), then validate at increasing depth — validate_node in minimal mode for required fields, full mode with profiles, then validate_workflow for connections and expressions. Deployment uses diff-based partial updates (n8n_update_partial_workflow) so the agent doesn't resend whole workflow JSON on every change.
Deployment options run the gamut: npx over stdio, Docker (ghcr.io/czlonkowski/n8n-mcp), Railway one-click, an HTTP mode for remote hosting, and the hosted dashboard.n8n-mcp.com if you don't want infrastructure. There are companion Claude skills (n8n-skills repo) and setup docs for Claude Code, Cursor, Windsurf, VS Code, and Codex.
Trade-offs: the node database is pinned to an n8n release (currently n8n 2.27.4) and updated regularly, so very new nodes can lag. Management tools are exactly as dangerous as the API key you hand them — the read-only deployment recipe (DISABLED_TOOLS + read-only key) exists for a reason. Telemetry is collected with a documented opt-out (PRIVACY.md).
Tools & capabilities (22)
tools_documentationGet usage documentation for any n8n-MCP tool — the recommended first call
search_nodesFull-text search across 2,063 core and community nodes, optionally with real config examples
get_nodeUnified node info: properties at minimal/standard/full detail, docs mode, property search, version diffs
validate_nodeValidate a node configuration — quick required-fields check or full profile-based validation
validate_workflowValidate a complete workflow including connections, expressions, and AI Agent wiring
search_templatesSearch 2,352 workflow templates by keyword, node types, task, or metadata (complexity, audience)
get_templateFetch a template's full workflow JSON (nodes_only / structure / full modes)
n8n_create_workflowCreate a new workflow on your n8n instance (requires API key)
n8n_get_workflowRetrieve a workflow by ID at several detail levels
n8n_update_full_workflowReplace an entire workflow
n8n_update_partial_workflowUpdate a workflow with token-efficient diff operations
n8n_delete_workflowPermanently delete a workflow
n8n_list_workflowsList workflows with filtering and pagination
n8n_validate_workflowValidate a deployed workflow by ID
n8n_autofix_workflowAutomatically fix common workflow errors
n8n_workflow_versionsManage workflow version history and rollback
n8n_deploy_templateDeploy a template from n8n.io straight to your instance with auto-fix
n8n_test_workflowTrigger a workflow execution (webhook, form, chat) for testing
n8n_executionsList, inspect, and delete workflow executions
n8n_manage_credentialsList, create, update, delete n8n credentials and fetch credential schemas
n8n_audit_instanceSecurity audit combining n8n's audit API with deep workflow scanning
n8n_health_checkCheck n8n API connectivity and available features
What this server can do
n8n-MCP provides tools for these capabilities — tap one to see every MCP server that does the same:
When to use it
- Build an n8n workflow from a plain-language prompt, with node configs validated before import
- Search 2,352 templates by task or complexity and deploy one directly to your instance
- Diagnose and auto-fix a broken workflow: validate connections and expressions, then n8n_autofix_workflow
- Monitor executions and trigger test runs from the chat instead of the n8n UI
Quick setup
- 1Add the npx config to your MCP client — with no env vars you get the 7 documentation/validation tools
- 2Optional: create an API key in your n8n instance (Settings → API) and add N8N_API_URL + N8N_API_KEY to unlock management tools
- 3Restart the client and start with a call to tools_documentation()
- 4Docker alternative: docker run -i --rm ghcr.io/czlonkowski/n8n-mcp:latest with the same env vars
- 5No-install alternative: sign up at dashboard.n8n-mcp.com (free tier, 100 calls/day)
Security notes
Documentation tools run locally against a bundled database and need no credentials. With an N8N_API_KEY configured, the server can create, modify and delete workflows on your instance — the README explicitly warns never to let AI edit production workflows without copies and backups. For governance-sensitive setups, combine a read-only n8n API key with the DISABLED_TOOLS env var.
n8n-MCP FAQ
Do I need a running n8n instance to use n8n-mcp?
No. Node search, documentation, templates and validation work standalone from a bundled database. You only need an n8n instance plus API key for the 15 management tools that create, update and execute workflows.
Is n8n-mcp free?
Yes — MIT-licensed and free to self-host via npx or Docker. The optional hosted service at dashboard.n8n-mcp.com has a free tier of 100 tool calls/day.
Is this the official n8n MCP server?
No. It's a community project by Romuald Czlonkowski, not n8n GmbH — though at 22k+ GitHub stars it's the de facto standard way to connect AI assistants to n8n. n8n itself also ships an MCP Client Tool node that can consume this server.
Alternatives to n8n-MCP
Compare all alternatives →
AI-powered task-management system for AI-driven development that drops into Cursor, Windsurf, Claude Code, and more.
Background computer-use MCP server that drives native macOS, Windows, and Linux desktop apps without stealing focus.
Self-hosted MCP server for Jira and Confluence Cloud and Server/Data Center.