n8n is an open source AI agentlow-code AI Agent and workflow automation platform, connecting 500+ services and AI models through a visual node editor, supporting MCP protocol, self-hosted deployment and fair code licensing.

n8n: Open Source AI Agent and Workflow Automation Platform

n8n’s core parameters and statistics

Parameter	Value
Product positioning	Open source low-code AI Agent and workflow automation platform (iPaaS)
Core form	Web visual editor + API + Docker self-hosting + cloud hosting
Target users	DevOps, IT teams, business operations AI developers, automation engineers
Core technology	Visual node orchestration AI Agent engine MCP protocol, pluggable node system
License Agreement	Sustainable Use License (Fair Code) + Enterprise License
Deployment method	Docker self-hosting / n8n Cloud / Enterprise private cloud
Number of integrations	500+ nodes (including 1500+ community-contributed integrations)
Latest stable version	2.30.8 (2026-07-20)
GitHub Stars	197k+
GitHub Forks	59.5k+
Community members	200k+
G2 Rating	4.7/5

The differentiation of n8n lies in the triple combination of "open source + self-hosting + native support for AI Agent". Compared with closed source solutions such as Zapier and Make, n8n allows the complete workflow engine to be deployed on the enterprise's own infrastructure, and the data does not leave the network boundary. Version 2.0, launched in 2025, will upgrade the product positioning from "low-code workflow automation" to "AI Agent and workflow automation platform", adding key capabilities such as AI Agent node MCP protocol integration and AI workflow builder.

n8n’s users and market recognition

GitHub Open Source Ecosystem: n8n is one of the most popular workflow automation projects on GitHub, with 197k+ Stars ranking in the Top 50 global open source projects. 59.5k Forks and 775+ contributors indicate that the community is extremely active, with an average of multiple code submissions and issue responses per day. 727 official Releases cover the complete iteration path from 0.x to 2.30.x.

Enterprise Customers: Officially announced cases include food brand Huel (saving 1,000 hours of labor), Vodafone (saving £2.2 million through threat intelligence automation), SanctifAI (AI workflow development 3 times faster than Python solutions), etc. These cases cover multiple industries such as retail and telecom AI services.

G2 Rating: 4.7/5 stars. Keywords often mentioned by users are "visualization + code dual mode", "self-hosted flexibility" and "rich community ecology". The community forum (community.n8n.io) has 200k+ registered members and a large number of templates to share.

Market Positioning: In the field of open source iPaaS, n8n is currently the solution with the most complete functions and the most active community. Products that compete with it include Zapier (closed source), Make (closed source), Activepieces (open source but with a smaller community), and Dify (focused on AI applications rather than general workflows).

The cost advantage of n8n

n8n’s cost structure needs to be understood at three levels:

• C-side/Self-hosted (Community Edition): completely free. The Sustainable Use License allows individuals and small teams (with annual income below the threshold) to use all features for free, including AI node MCP integration and code nodes. The functions are the same as the Cloud version without any emasculation. The hidden cost is the need to maintain your own servers, databases and backup strategies.

• Developer/API call: There is no API call fee in self-hosted mode, and only consumes own server resources. The Cloud version is billed by the number of workflow executions, not by the number of API calls - each workflow is counted as one execution from start to finish, with no limit on the number of steps. This is a key difference from other platforms (which charge per step).

• Business/Private Cloud (Business/Enterprise): Business plan €667/month (annual payment), including 40k executions/month SSO/SAML/LDAP, version control (Git), and multi-context support. Contact sales for the Enterprise plan for customization, including unlimited projects, 200+ concurrent executions, 365-day execution log retention, and exclusive SLA. Both plans support self-hosted deployments.

Cost comparison with other platforms:

Compare Dimensions	n8n (self-hosted)	Zapier	Make
Basic fee	Free	Starting from $19.99/month	Starting from €9/month
Billing unit	By number of executions	By tasks/steps	By number of operations
Self-Hosted	✅ Supported	❌ Not Supported	❌ Not Supported
AI Node	Free built-in	Additional payment	Partially paid
User limit	Unlimited	By package	By package
Data control	Complete self-control	Platform hosting	Platform hosting

Hidden Cost Tip: Self-hosting requires investment in operation and maintenance manpower (Docker management, database backup, version upgrade). For small teams without full-time DevOps, the cost of this part of time may exceed the subscription fee of the Cloud version. It is recommended that small teams first try the Cloud free trial period and confirm the stability of the workflow before deciding on the deployment method.

Main functions of n8n

• AI Agent native support: core upgrade of n8n 2.0. Create AI Agent nodes directly in the workflow, configure system prompt words, select underlying LLM (OpenAI, Anthropic, Google Gemini, open source models, etc.), and bind tools (Tools). Agent can call other nodes in the workflow as tools to realize the "thinking + action" combination. Supports multi-Agent collaborative orchestration and RAG retrieval enhanced generation.

• MCP (Model Context Protocol) integration: n8n instances can be used as MCP Server to expose workflows, allowing external AI platforms such as Claude and Lovable to directly call n8n automation processes. At the same time, n8n workflow can also call external MCP tools to open a two-way channel between the AI ​​platform and business systems. This is the key differentiator between n8n and traditional workflow engines.

• 500+ integration nodes: covering mainstream SaaS (Google Workspace, Slack, Salesforce, HubSpot), databases (PostgreSQL, MySQL, MongoDB, Redis), cloud services (AWS, Azure, GCP), AI models (OpenAI, Anthropic, Hugging Face, Ollama), communications (Email, SMS, Telegram, WhatsApp), etc. The total number of community-contributed nodes exceeds 1,500.

• AI Workflow Builder (Natural Language → Workflow): Experimental feature coming in 2025. Users describe automation requirements in natural language, and AI automatically generates the corresponding workflow structure. For example, if you enter "Pull yesterday's order data from PostgreSQL at 9 a.m. every day, calculate and summarize it and send a Slack notification to the sales group", n8n will automatically create a complete process of scheduled triggering + database query + data aggregation + Slack message.

• Visual workflow editor + code node: Drag-and-drop node editing, and data is transferred between nodes through connections. Each node displays a live input and output preview. When visualization nodes cannot meet your needs, you can insert JavaScript or Python code nodes to directly manipulate data, call npm packages, or execute complex logic. The input and output of code nodes are also visible in the editor.

• Human-in-the-Loop (human-computer collaboration): Insert a human approval node anywhere in the workflow. The AI ​​Agent will pause when it reaches this node and wait for manual confirmation/modification before continuing. Approval requests can be reached to approvers through Slack, email, web interface, etc. This mechanism is a key function for AI automation to be implemented in production environments, especially in high-value or irreversible operation scenarios.

• Chat Hub (Enterprise AI Chat Hub): A unified AI conversation interface that allows team members to access multiple AI models and deployed Agent workflows in a single chat window. Administrators can configure the range of AI tools and models available to different roles to achieve enterprise-level AI access control.

• Conditional branching, looping and error handling: Supports process control such as if/else, switch multi-way branch Loop Over Items (batch looping), Wait and Try/Catch error capture. Loop nodes can automatically process API paging data, pulling the next page with each iteration until the data is exhausted.

• Webhook and multiple triggering methods: Supports Webhook URL to receive external events, Cron scheduled scheduling, polling detection, manual triggering, event-driven triggering, etc. Each workflow can be configured with multiple triggers to meet different starting conditions.

n8n model and version evolution

The version evolution of n8n can be roughly divided into four stages:

Early exploration period (0.x, ~2021-2024)

• 0.x series: The initial version of n8n, the core value is "open source alternative to Zapier". Established visual node editor infrastructure and pluggable node system. The number of nodes has grown from dozens to 200+. The license agreement was originally Apache 2.0 and later changed to Sustainable Use License.

Maturity period of enterprise (1.x LTS, ~2024-2025)

• 1.0.0 (~2024-06): The first LTS mainline version, introducing a new orchestration engine, node version management, variable system, and error workflow. New enterprise-level features: SSO/SAML/LDAP, RBAC permission model, audit log. The node ecosystem has expanded to 400+. The 1.x series is under ongoing maintenance to this day (currently 1.123.66), receiving security fixes as an LTS branch.

AI native transformation period (2.0, ~2025-2026)

• 2.0.0 (~2025-06): The product positioning is upgraded from "workflow automation" to "AI Agent and workflow automation platform". Core changes include: native AI Agent node (supports multi-model Tool Calling, RAG), MCP protocol integration (bidirectional), AI workflow builder (natural language → process), and Chat Hub chat center. This release redefines n8n’s position in the automation market.

Continuous iteration period (2.30.x, 2026)

• 2.30.x series (active 2026-07): Current stable branch, latest 2.30.8 (2026-07-20). Continuously optimize AI Agent stability, MCP execution mode, HITL tool connection, etc. The 2.31.x pre-release branch is maintained at the same time for new feature verification.

Version line	Release date	Key changes	Current status
0.x	~2021	Initial open source version, 200+ nodes	Maintenance has stopped
1.x LTS	~2024-06	Enterprise-level functions 400+ nodes, orchestration engine reconstruction	Security maintenance (1.123.66)
2.0	~2025-06	AI Agent natively supports MCP integrated AI builder	Current mainline (2.30.8)
2.31.x	~2026-07	New feature verification branch	Pre-release

n8n’s technical advantages

"Visual + Code" dual-mode architecture: n8n does not force users to choose between "low code" and "pure code". Visual nodes handle 80% of standard integration scenarios, and code nodes handle 20% of edge customization needs. The two are mixed in the same workflow, and the data flow is fully visible on the same canvas. This design lowers the threshold for team collaboration—business personnel can build a basic process framework, and developers can insert code logic when needed.

The working mechanism of the AI ​​Agent engine: n8n's AI Agent node is essentially a circular reasoning engine - the Agent receives user input → determines whether a Tool needs to be called → generates a Tool Call if necessary → executes the corresponding node → feeds the results back to the LLM → continues reasoning until the final response is generated. All reasoning steps (Prompt, Tool Call, model reply) are completely recorded in the workflow execution log, which can be traced and audited. This "transparent AI" design is a key requirement for enterprise compliance scenarios.

MCP Bidirectional Integration: n8n is both an MCP Client and an MCP Server. As a Server, it exposes the n8n workflow as a standard MCP Tool for clients such as Claude and Lovable to call; as a Client, it can call external MCP Servers in the workflow. This two-way design makes n8n a "connection hub" between the AI ​​platform and existing business systems.

Reliability design of execution engine: n8n's execution engine supports production-level features such as queue mode (multi-instance horizontal expansion), automatic retry, error workflow (Workflow-level try/catch), and execution log persistence. Workflow executions are recorded as immutable events, supporting traceback playback and debugging. For Cloud edition, n8n offers an availability SLA of 99.9%+.

Self-hosted security boundary: In self-hosted mode, all data (credentials, execution log AI requests) flow within the enterprise network. n8n supports encrypted credential storage, external key management integration (such as HashiCorp Vault), SAML/LDAP SSO, forced 2FA, RBAC permission model. For enterprises with strict compliance requirements such as SOC2 and GDPR, self-hosting is an unavoidable prerequisite.

How to use n8n

n8n provides three usage methods:

Method	Applicable Scenario	Startup Command/Entrance
Self-hosting (Docker)	Sensitive data, high compliance requirements, large-scale production environment	docker run -it --rm --name n8n -p 5678:5678 -v n8n_data:/home/node/.n8n docker.n8n.io/n8nio/n8n
Self-hosted (npm)	Quick trial, development testing	npx n8n (requires Node.js)
n8n Cloud	No operation and maintenance required, quick to get started, small team	Register https://app.n8n.cloud/register

Typical getting started process:

1. Deployment/Registration: Choose self-hosted (Docker) or sign up for a Cloud free trial (14 days, no credit card required).
2. Create a Workflow: Start from blank or choose from a library of 9000+ templates. The template covers AI Agent, data synchronization, notification integration, content processing and other scenarios.
3. Configure triggers: Set workflow startup conditions - Webhook URL, Cron timing, event monitoring or manual triggering.
4. Drag-and-drop arrangement: Select the integrated node from the node panel and drag it to the canvas, and define the data flow sequence through connections. The configuration form of each node displays input/output samples in real time.
5. Add AI node: Insert the AI ​​Agent or LLM node, configure the model connection (API Key), system prompt words and Tool binding.
6. Testing and Debugging: Click "Execute Workflow" to run the complete process and view the input and output of each node. Supports single-node rerun without re-executing the entire workflow.
7. Deployment goes online: After confirming that everything is correct, click "Save" and activate the workflow. It is recommended for production environments to enable execution logging and error notifications.

AI Agent Workflow Quick Example: Create an "Intelligent Customer Service Work Order Classification" Agent - configure Webhook triggers to receive work orders → AI Agent analyzes the work order content → Branch routing to the corresponding processing node (refund/technical support/complaint) → Update the work order system → Slack notifies the person in charge. The entire process is configured in the visual editor without writing any code.

n8n product pricing

n8n adopts a dual-track pricing model of "open source core + cloud service value-added":

Plan	Price	Number of executions	Concurrency	AI quota	Key functions
Community (self-hosted)	Free	Unlimited	Unlimited	Unlimited	All core functions, code node MCP
Starter (Cloud)	€20/month (paid annually)	2,500/month	5	2,300/month	1 shared project, forum support
Pro (Cloud)	€50/month (paid annually)	10,000/month	20	Up to 13,700/month	7 days for 3 projects Insights, Git version control
Business (Self-hosted/Cloud)	€667/month (paid annually)	40,000/month	200+	On-demand	6 projects SSO/SAML, 30 days Insights, AI Assistant (coming soon), Duoyoujing
Enterprise	Contact Sales	Custom	200+	On-demand	Unlimited projects 365-day logs, dedicated SLA, log streaming, external key storage

Billing mechanism highlights: n8n charges based on "complete workflow execution" instead of competitors' "based on steps/operations". A complex 20-step workflow consumes execution quota only once per run. This can significantly reduce costs in mass production scenarios.

AI Assistant Credits: The AI ​​workflow builder introduced in n8n 2.0 consumes AI Credits (for natural language → workflow conversion). Starter plan includes 2,300/month, Pro plan up to 13,700/month, Business/Enterprise configuration on demand. This feature remains in preview as of July 2026 and pricing is subject to adjustment.

Startup Program: Startups with < 20 employees can apply for a 50% discount on the Business plan.

Free mode limitations: The self-hosted Community version is fully functional and non-censored, but officially does not provide SLA and technical support (only community forum). For production-critical workflows, Business and above plans are a safer choice.

Application scenarios of n8n

• AI-driven customer service ticket processing: Enterprises embed AI Agent into the customer service process. Work orders enter n8n through Webhook → AI Agent automatically classifies, extracts key information, and matches the knowledge base → automatically responds or assigns to human customer service based on confidence → updates CRM and work order systems. Cost reduction and efficiency increase deduction: For a customer service team with an average of 500 work orders per day, AI preprocessing can take over about 60% of regular inquiries, shortening the manual response time from "4 hours" to "within 5 minutes", which is equivalent to saving the labor cost of 2-3 full-time customer service personnel.

• Data ETL and business synchronization: regularly pull order data from ERP/e-commerce platform → clean and convert → synchronize to data warehouse CRM and financial system. Cost reduction and efficiency improvement: In the past, data engineers were required to write and maintain Python ETL scripts (about 2 days/time). Using n8n visual configuration, the configuration can be completed within 2 hours, and subsequent changes do not require development intervention. This scenario is suitable for 100% automation, requiring only manual intervention to correct abnormal data.

• AI content production pipeline: The operations team builds an automated content pipeline - RSS monitoring of competitive product dynamics → AI summary + translation → manual review (HITL) → automatic publishing to CMS and social media platforms. Multiple AI nodes can be used in combination: one node does content classification, one does SEO keyword extraction, and one does multi-language translation.

• IT Operation and Maintenance Automation: Monitor alarm access to n8n Webhook → Automatically create Jira work orders → AI Agent analyzes alarm logs and recommends repair solutions → Automatically execute recovery scripts (such as restarting services) → Notify on-duty engineers in case of failure. Cost reduction and efficiency increase deduction: In the case of Vodafone, n8n’s automated process for processing security threat intelligence saved £2.2 million per year.

• Multi-Agent collaborative research: The marketing team configures an "AI analyst team" - one Agent is responsible for market trend search, one Agent is responsible for competitive product analysis, and one Agent is responsible for report writing, and collaboratively completes in-depth research reports through workflow orchestration. Results are automatically pushed to team groups via WhatsApp/Email.

Scenario Adaptation Boundary: n8n is best suited for cross-system integration and AI-enhanced process automation. Unsuitable scenarios include: real-time requirements for sub-second transaction processing (an event streaming platform such as Kafka should be used), very large-scale data processing (using a dedicated ETL engine such as Airbyte), and pure AI chatbots (using Dify or Botpress to be more lightweight).

n8n Applicable people

• DevOps/Platform Engineer: n8n is the ideal tool for "automating glue code". It is suitable to replace handwritten glue code (Python script Shell pipeline) with visual workflow to reduce the maintenance cost of operation and maintenance scripts. Prerequisites: Requires knowledge of Docker and basic networking from managed deployments.

• AI application developer: Suitable for quickly building AI Agent prototypes and deploying them to production environments. n8n's AI Agent node abstracts complex details such as LLM calling Tool Calling and RAG, allowing developers to focus on the Agent logic design rather than the infrastructure. Prerequisite: Understand the basic concepts of Prompt Engineering.

• Business Operations Staff (with technical background): Suitable for handling cross-system data synchronization and notification automation. No programming experience is required to build most integration scenarios. But you need to understand basic concepts such as API authentication JSON data structure. For users with zero programming foundation, it is recommended to complete the n8n official introductory course first.

• Enterprise IT Managers: For teams evaluating "low-code automation platforms". n8n’s self-hosting capabilities and Sustainable Use License make it compliantly feasible for large enterprises. It is recommended to pilot 1-2 business scenarios with a small team first, and then evaluate the feasibility of expanding to the entire company.

Not applicable to the crowd:

• Business personnel who need completely zero code (just drag and drop, no technical concepts involved) - Although n8n provides visual editing, configuring API credentials, processing JSON data, and debugging workflows still require a certain technical foundation.
• Requires out-of-the-box, maintenance-free automated services - Self-hosting requires operation and maintenance investment, and although the Cloud version eliminates operation and maintenance, there is still a learning cost.
• Very large enterprises (1000+ employees) with strict restrictions on licensing compliance - the annual revenue threshold provisions of the Sustainable Use License require legal evaluation.

n8n Summary and Outlook

The core competitiveness of n8n lies in the trinity positioning of "open source + native AI agent + self-hosting". In the early 2020s, it was just an open source alternative to Zapier, and after version 2.0 in 2025, it has transformed into an AI Agent workflow platform. In the low-code automation market, n8n is currently the only solution that meets the following four conditions at the same time: it is completely open source, supports self-hosting, has a built-in AI Agent engine, and has enterprise-level security and governance capabilities.

Current Limitations and Uncertainties:

• The income threshold clause of the Sustainable Use License has a gray area for compliance by large enterprises. Enterprises with annual income exceeding the threshold need to purchase a commercial license.
• AI Agent nodes (2.0 core functions) and AI workflow builder are still in a rapid iteration period, and production-bound use requires attention to version upgrade compatibility.
• The readability of the canvas for complex workflows (50+ nodes, multi-level branch nesting) is reduced, and the graphical debugging tools are not as intuitive as the code IDE.
• Performance optimization in high-concurrency scenarios (thousands of workflows running simultaneously) mainly relies on queue mode and horizontal expansion, and requires reasonable architectural design.

Follow-up observation points:

1. AI Assistant’s pricing strategy from preview to GA – will affect the total cost of ownership of n8n Cloud.
2. MCP ecological development - whether n8n’s pioneering advantage in MCP Server/Client two-way integration can be transformed into network effects.
3. Localization for the Chinese market - Currently there are no Chinese interfaces and domestic cloud deployment options, and there are usage barriers for domestic users.
4. Maturity of Agent evaluation (Evals) function - this is a key capability for AI Agent to move from demo to production.

Procurement/Adoption Risk Assessment: It is recommended that small teams (<20 people) start with Community Edition self-hosted, 0 cost to verify that n8n's node coverage and workflow stability meet business needs. After confirming that it is suitable, you can upgrade to the Cloud Pro plan (€50/month) if you need to reduce the operation and maintenance burden. Medium-sized enterprises (20-100 people) are recommended to directly adopt the Business plan (€667/month) to establish a standardized automated governance system using SSO, Git version control and multi-context support. Large enterprises (100+ people) need to confirm with their legal department the applicable conditions for the annual revenue threshold of the Sustainable Use License before purchasing, and evaluate whether the customized SLA and dedicated support of the Enterprise plan meet compliance requirements. You can refer to the public cases of Vodafone, Huel and other similar companies as a reference for decision-making.

Version Info

• n8n 2.30.8 ：Stable version, fixes the problem of AI Agent connection through HITL tool, automatic ungrouping of MCP workflow execution mode requirements when saving workflow, etc. （2026-07-20）
• n8n 2.0 ：There is no official precise date yet. AI Agent is introduced to natively support MCP protocol, integrate AI workflow builder (natural language creation workflow), and Chat Hub multi-model chat interface. （~2025-06）
• n8n 1.x LTS ：There is no official precise date for the introduction of new orchestration engines, node version management, enterprise-level SSO and RBAC, becoming the main line of long-term support. （~2024-06）
• n8n 0.x ：There is no official precise date yet. An early version of n8n, establishing a basic automated node ecosystem and visual editor architecture. （~2021）