Provides structured access to n8n nodes, properties, operations, and templates, enabling AI agents to search, validate, and manage n8n workflows.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@n8n MCP Server find a node for sending Slack messages"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Servidor MCP (Model Context Protocol) para n8n — fork de czlonkowski/n8n-mcp con modificaciones para despliegue en producción sobre Contabo VPS.
Este servidor actúa como puente entre n8n y asistentes de IA (Claude, Hermes Agent, etc.), proporcionando acceso estructurado a la documentación de nodos, propiedades, operaciones y plantillas de n8n.
Related MCP server: n8n-MCP
🔌 Servidor MCP completo — expone herramientas para buscar, validar y gestionar workflows de n8n
📚 Documentación de nodos — acceso a 1,845+ nodos (816 core + 1,029 community)
🐳 Dockerizado — despliegue con un solo comando usando Docker Compose
🔐 Autenticación por token — protegido con AUTH_TOKEN
🌐 Modo HTTP — compatible con cualquier cliente MCP que soporte SSE (Server-Sent Events)
🏗️ Multi-tenant opcional — soporte para múltiples instancias n8n
📊 Telemetría anónima desactivada por defecto en este fork
Docker y Docker Compose instalados
Una instancia de n8n corriendo (local o remota)
(Opcional) API Key de n8n para herramientas de gestión de workflows
git clone <url-de-este-repo> n8n-mcp
cd n8n-mcpCopia el archivo de ejemplo y edita las variables:
cp .env.example .envEdita .env con tus valores reales. Las variables mínimas requeridas son:
AUTH_TOKEN=*** 3. Construir y levantar
```bash
docker compose up -d --buildEl servidor estará disponible en http://localhost:3001 (mapeado a 127.0.0.1:3001 por seguridad).
curl http://localhost:3001/health
# Respuesta esperada: {"status":"ok"}El servicio se conecta a la red baota_net (externa), que debe existir previamente. Esta es la misma red donde corre n8n, permitiendo comunicación interna via http://n8n:5678.
Si tu configuración de red es diferente, edita la sección networks en docker-compose.yml.
Variable | Descripción | Valor por Defecto |
| Modo del servidor ( |
|
| Puerto HTTP |
|
| Host de escucha |
|
| Entorno ( |
|
| Nivel de logging |
|
| Ruta a la base de datos SQLite |
|
| Reconstruir BD al iniciar |
|
Variable | Descripción | Requerido |
| Token de autenticación para HTTP | Sí |
| URL pública del servidor | No |
| Modo de protección SSRF ( |
|
| Confiar en proxy inverso (0 o 1) |
|
Variable | Descripción | Requerido |
| URL de la instancia n8n (sin | Sí |
| API Key de n8n | Sí |
| Timeout de requests (ms) |
|
| Reintentos máximos |
|
Variable | Descripción | Valor por Defecto |
| Desactivar telemetría |
|
Variable | Descripción | Valor por Defecto |
| Activar modo multi-tenant |
|
| Estrategia de sesiones ( |
|
Ve a tu instancia n8n → Settings → API
Crea una nueva API Key
Copia la key generada
.envN8N_API_URL=http://n8n:5678
N8N_API_KEY=*** **Nota:** Si n8n corre en el mismo Docker network (`baota_net`), usa el nombre del contenedor (`n8n`) como hostname. Si está en otro servidor, usa la URL completa.
### 3. Verificar conectividad
```bash
curl -H "Authorization: Bearer *** http://localhost:3001/healthPara conectar este servidor MCP a Hermes Agent (by Nous Research):
docker compose ps
# n8n-mcp debe estar "Up"En la configuración de Hermes Agent, agrega un nuevo servidor MCP:
{
"mcpServers": {
"n8n-mcp": {
"type": "http",
"url": "https://tu-dominio.com/mcp",
"headers": {
"Authorization": "Bearer <AUTH_TOKEN>"
}
}
}
}O si estás en la misma red local:
{
"mcpServers": {
"n8n-mcp": {
"type": "http",
"url": "http://localhost:3001/mcp",
"headers": {
"Authorization": "Bearer <AUTH_TOKEN>"
}
}
}
}Hermes Agent debería detectar automáticamente las herramientas disponibles. Puedes verificar preguntando:
"¿Qué herramientas de n8n tienes disponibles?"
Este repositorio es un fork de czlonkowski/n8n-mcp. Para actualizar con los últimos cambios:
git remote add upstream https://github.com/czlonkowski/n8n-mcp.gitgit fetch upstream
git checkout main
git merge upstream/mainSi hay conflictos (especialmente en docker-compose.yml, .env.example, o .gitignore), resuélvelos manualmente preservando las configuraciones locales de producción.
docker compose down
docker compose up -d --builddocker compose logs -f n8n-mcp
curl http://localhost:3001/healthn8n-mcp/
├── src/
│ ├── mcp/ # Servidor MCP y herramientas
│ ├── services/ # Validación, templates, ejemplos
│ ├── database/ # Capa de acceso a datos (SQLite)
│ ├── telemetry/ # Sistema de telemetría (desactivado)
│ └── utils/ # Utilidades
├── data/
│ ├── skills/ # Documentación de habilidades para IA
│ └── nodes.db # Base de datos de nodos (generada)
├── scripts/ # Scripts de utilidad y testing
├── tests/ # Tests unitarios e integración
├── docker/ # Archivos de configuración Docker
├── docker-compose.yml # Configuración de despliegue
├── Dockerfile # Build multi-etapa optimizado
├── .env.example # Plantilla de variables de entorno
└── package.json # Dependencias y scriptsNUNCA subas el archivo .env al repositorio — está en .gitignore
Usa AUTH_TOKEN fuerte (generar con: openssl rand -base64 32)
El puerto está mapeado a 127.0.0.1:3001 — solo accesible localmente
Usa un proxy inverso (Nginx/Traefik) con HTTPS para exponerlo públicamente
La telemetría está desactivada por defecto (N8N_MCP_TELEMETRY_DISABLED=true)
El modo SSRF está en permissive para comunicación interna — ajustar según necesidades
Este repositorio ha sido auditado para garantizar que no contiene:
API keys reales
Tokens JWT completos
Credenciales de Supabase
Contraseñas o secrets de producción
Todos los valores en archivos de test son placeholders truncados (ej: eyJhbG...nkCk, test123...890).
MIT — ver archivo LICENSE.
Basado en czlonkowski/n8n-mcp por Romuald Czlonkowski.
Get node info with progressive detail levels and multiple modes. Detail: minimal (~200 tokens), standard (~1-2K, default), full (~3-8K). Modes: info (default), docs (markdown documentation), search_properties (find properties), versions/compare/breaking/migrations (version info). Use format='docs' for readable documentation, mode='search_properties' with propertyQuery for finding specific fields.
Inputs: nodeType (required), mode, detail, toVersion, +5 more.
Get template by ID. Use mode to control response size: nodes_only (minimal), structure (nodes+connections), full (complete workflow).
Inputs: templateId (required), mode.
Security audit of n8n instance. Combines n8n's built-in audit API (credentials, database, nodes, instance, filesystem risks) with deep workflow scanning (hardcoded secrets via 50+ regex patterns, unauthenticated webhooks, error handling gaps, data retention risks). Returns actionable markdown report with remediation steps using n8n_manage_credentials and n8n_update_partial_workflow.
Inputs: categories, customChecks, includeCustomScan, daysAbandonedWorkflow.
Automatically fix common workflow validation errors. Preview fixes or apply them. Fixes expression format, typeVersion, error output config, webhook paths, connection structure issues (numeric keys, invalid types, ID-to-name, duplicates, out-of-bounds indices).
Inputs: id (required), fixTypes, maxFixes, applyFixes, +1 more.
Create workflow. Requires: name, nodes[], connections{}. Created inactive. Returns workflow with ID.
Inputs: name (required), nodes (required), connections (required), settings, +1 more.
Permanently delete a workflow. This action cannot be undone.
Inputs: id (required).
Deploy a workflow template from n8n.io directly to your n8n instance. Deploys first, then auto-fixes common issues (expression format, typeVersions). Returns workflow ID, required credentials, and fixes applied.
Inputs: templateId (required), name, autoFix, stripCredentials, +1 more.
Manage workflow executions: get details, list, or delete. Use action='get' with id for execution details, action='list' for listing executions, action='delete' to remove execution record.
Inputs: action (required), id, mode, limit, +12 more.
Generate an n8n workflow from a natural language description using AI. Call with just a description to get workflow proposals. Then call again with deploy_id to deploy a chosen proposal, or set skip_cache=true to generate a fresh workflow. Use confirm_deploy=true to deploy a previously generated workflow.
Inputs: description (required), deploy_id, skip_cache, confirm_deploy.
Get a single node's full configuration from a workflow by node name. Returns parameters, type, position, and other metadata without downloading the entire workflow. Use this instead of n8n_get_workflow(mode="full") when you only need one node's config. Use mode='filtered' with nodeNames for fetching multiple specific nodes.
Inputs: nodeName (required), workflowId (required), fields.
Get workflow by ID with different detail levels. n8n has a draft/publish model: the workflow body holds the draft (latest edits); use mode='active' to see the published graph that is actually running. Modes: 'full' (draft + metadata), 'details' (full + execution stats), 'active' (published graph only), 'structure' (nodes/connections topology), 'filtered' (full config of only the nodes named in nodeNames - use to read one heavy node without the whole workflow), 'minimal' (id/name/active/tags).
Inputs: id (required), mode, nodeNames.
Check n8n instance health and API connectivity. Use mode='diagnostic' for detailed troubleshooting with env vars and tool status.
Inputs: mode, verbose.
List workflows (minimal metadata only). Returns id/name/active/dates/tags. Check hasMore/nextCursor for pagination.
Inputs: tags, limit, active, cursor, +2 more.
Manage n8n credentials. Actions: list, get, create, update, delete, getSchema. Use getSchema to discover required fields before creating. For list, page beyond 100 results with cursor (from the previous response's nextCursor). NOTE: list/get need an n8n deployment whose public API permits credential reads — older n8n versions, restricted API keys, or instance settings can reject them, returning NOT_SUPPORTED (create, delete, getSchema — and update where the API version supports it — still work). SECURITY: credential data values are never logged.
Inputs: action (required), id, data, name, +4 more.
Manage n8n data tables and rows. Actions: createTable, listTables, getTable, updateTable, deleteTable, getRows, insertRows, updateRows, upsertRows, deleteRows.
Inputs: action (required), data, name, limit, +10 more.
Read the current value of a specific field from a workflow node without downloading the full workflow. Use this to inspect a field before editing it with n8n_update_partial_workflow (patchNodeField).
Inputs: nodeName (required), fieldPath (required), workflowId (required).
Test/trigger workflow execution. Auto-detects trigger type (webhook/form/chat). Supports: webhook (HTTP), form (fields), chat (message). Note: Only workflows with these trigger types can be executed externally.
Inputs: workflowId (required), data, headers, message, +6 more.
Full workflow update. Requires complete nodes[] and connections{}. For incremental use n8n_update_partial_workflow.
Inputs: id (required), name, nodes, settings, +1 more.
Update workflow incrementally with diff operations. Types: addNode, removeNode, updateNode, patchNodeField, moveNode, enable/disableNode, addConnection, removeConnection, updateSettings, updateName, add/removeTag, activate/deactivateWorkflow, transferWorkflow. patchNodeField requires fieldPath (dot path, e.g. "parameters.jsCode") and patches: [{find, replace}]. See tools_documentation("n8n_update_partial_workflow", "full") for details.
Inputs: id (required), operations (required), validateOnly, continueOnError.
Validate workflow by ID. Checks nodes, connections, expressions. Returns errors/warnings/suggestions.
Inputs: id (required), options.
Manage workflow version history, rollback, and cleanup. Versions are scoped to your n8n instance. Five modes:
list: Show version history for a workflow
get: Get details of specific version
rollback: Restore workflow to previous version (creates backup first)
delete: Delete specific version or all versions for a workflow
prune: Manually trigger pruning to keep N most recent versions Old backups are also pruned automatically (10 most recent per workflow, plus an age-based retention window).
Inputs: mode (required), limit, deleteAll, versionId, +3 more.
Search n8n nodes by keyword with optional real-world examples. Pass query as string. Example: query="webhook" or query="database". Returns max 20 results. Use includeExamples=true to get top 2 template configs per node.
Inputs: query (required), mode, limit, source, +2 more.
Search templates with multiple modes. Use searchMode='keyword' for text search, 'by_nodes' to find templates using specific nodes, 'by_task' for curated task-based templates, 'by_metadata' for filtering by complexity/setup time/services, 'patterns' for lightweight workflow pattern summaries mined from 2700+ templates.
Inputs: task, limit, query, fields, +9 more.
Get documentation for n8n MCP tools. Call without parameters for quick start guide. Use topic parameter to get documentation for specific tools. Use depth='full' for comprehensive documentation.
Inputs: depth, topic.
Validate n8n node configuration. Use mode='full' for comprehensive validation with errors/warnings/suggestions, mode='minimal' for quick required fields check. Example: nodeType="nodes-base.slack", config={resource:"channel",operation:"create"}
Inputs: config (required), nodeType (required), mode, profile. Returns: valid, errors, summary, nodeType, +5 more
Full workflow validation: structure, connections, expressions, AI tools. Returns errors/warnings/fixes. Essential before deploy.
Inputs: workflow (required), options. Returns: valid, errors, summary, warnings, +1 more
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 26 tools
Most tools have distinct purposes, but validate_workflow and n8n_validate_workflow are very similar, and n8n_autofix_workflow overlaps with validation. This creates some ambiguity for an agent choosing which tool to use.
The majority of tools use a consistent 'n8n_verb_noun' pattern, but 7 tools lack the prefix (e.g., get_node, validate_workflow), and one tool (validate_node) has no corresponding prefixed version. This mixed convention is still readable but inconsistent.
With 26 tools, the count is on the higher side but justified by the broad domain coverage (workflows, credentials, templates, executions, validation, data tables). Each tool serves a clear purpose, and no tool feels extraneous.
The tool set covers CRUD for workflows, execution management, validation, fixing, versioning, credential and data table management, template deployment, and node configuration. Minor gaps exist (e.g., no explicit export/import), but the surface is largely complete for typical n8n management tasks.
Build, check and fix n8n workflows from Claude, Cursor or ChatGPT. Built on 1,100+ real n8n node definitions and curated templates, with a validator that repairs common mistakes (wrong node types, missing settings) and returns JSON ready to import into n8n. Sign in with a free gydmation.com account; nothing to install.
Static n8n workflow validator: workflow errors and repair suggestions. No execution.
Free public n8n, AI support and software acceptance templates; scoped private delivery workflows.
AI automation platform (n8n/Zapier alternative): build, schedule and run workflows from your agent.
