500+ready-made integrations to launch workflows without custom code

17LLM providers and 12 vector stores for RAG and AI agents out of the box

~200kactive users and 3000+ enterprise customers

~197kGitHub stars — a mature ecosystem and community

Integrations

Change one system without rewriting the rest

The integration layer cures four ailments of rigid data exchange: data loss, cascading rework, source overload and inconsistency. ESB, Kafka and n8n solve different problems.

50flows in 6 months — a speed benchmark for an ESB project

48flows in production with a target loosely coupled architecture

16xfaster launch of standard integrations compared to code

ESB

Routing, transformation, guaranteed delivery and low-code support of legacy exchanges.

Kafka

Durable log: an event is stored, re-read, multiple consumers read at their own pace.

n8n

Fast orchestration of a process and AI steps where heavy event streaming isn't needed.

sourcedata contractESB/Kafka/n8nmonitoringconsumers

Related Sections

ESB approach Kafka n8n Integration cases Apache Kafka News

Sources

Enterprise Integration Patterns Martin Fowler: Event-Driven Confluent: Kafka vs ESB

Industry solutions

What you can build with n8n

All solutions

E-commerce and retail Synchronizing orders, stock, and product records across marketplaces, warehouses, and ERP with AI-enriched descriptions Order and catalog management: less manual data transfer and fewer stock discrepanciesLearn more →Manufacturing Collecting events from systems (MES, 1C, sensors) and automatic alerts on failures and deviations Rapid response to production incidents without manual monitoringLearn more →Logistics Orchestrating delivery statuses across TMS, carriers and customer channels via webhooks Delivery transparency and automatic customer notifications at every stageLearn more →Finance and accounting Extracting data from invoices and documents through an AI step and posting it into the accounting system Processing incoming documents: less manual entry and fewer reconciliation errorsLearn more →Customer support AI classification and routing of requests from email, chats, and forms with auto-drafted replies First-pass ticket handling: faster replies and a lighter first lineLearn more →HR and recruiting Automating applicant screening, parsing resumes with an AI step, and updating the ATS Recruiting: faster initial candidate screeningLearn more →Marketing Collecting leads from forms and ad channels, enriching them and passing them to CRM on an event Lead management: no request lost between channel and CRMLearn more →B2B services and SaaS Connecting internal tools (Slack, Jira, databases) into end-to-end processes with AI summaries Internal operations: automating teams' routine cross-system tasksLearn more →

Capabilities

n8n capabilities

Trigger: webhook / schedule / system eventConnector node: reading data from CRM/ERP/emailAI step: classification, extraction, generation (LLM + RAG)Logic: branching, filters, custom code (JS/Python)Action: write to a system, send a notification, reply to a customerExecution log and monitoring; workflow export to Git

iPaaS workflow flow: an external event triggers a chain of nodes, an AI step processes the data, and the result flows back into business systems. It demonstrates loose coupling: the business logic lives in the workflow alongside the systems, and the core remains unchanged.

Visual workflows built from nodes

An integration can be built and read not only by a developer but by an analyst or product manager — reducing reliance on a single knowledge holder.

500+ Ready-to-Use Connectors

Connecting CRM, ERP, warehouses, messengers and databases takes days, not months: standard integrations aren't written from scratch.

Event-driven: triggers, webhooks, schedules

Processes run at the moment of an event (an order, an email, a system change) rather than on manual trigger — less delay and less manual work.

AI step inside the workflow

Request classification, document data extraction, and response generation are built into an existing process without a separate AI project.

Deterministic Execution Loop

Retries, queues, idempotency, logs, and graph versioning in Git make execution repeatable and observable, wrapping a probabilistic AI step in predictable runtime behavior.

Self-hosted deployment

Data, accounts, and LLM calls stay within your perimeter: the issue of sending corporate data to an external SaaS is removed, along with a compliance blocker for regulated industries. The tradeoff is that updates, backups, and recovery become your responsibility, so the operating model must be planned before launch.

Loose coupling of services

Business logic lives in workflows alongside the systems, not inside them: the core (e.g. ERP/CRM) is neither forked nor patched.

Custom code in a node (JS/Python)

A non-standard step is added precisely into the workflow itself — no need to spin up a separate microservice for a single transformation.

Workflow portability

A workflow exports as JSON and is versioned in Git — the process moves between teams and contractors without a rewrite.

Open-source and fair-code license

No limits on executions or integrations and no vendor lock-in: total cost of ownership stays predictable as load grows.

Approach

How we implement n8n

Without modifying the core

We don't fork or patch the n8n core. n8n stays on the standard, upgradable version — we move business logic into separate microservices alongside it, so platform updates don't break your customizations.

International Standards, Not Homegrown Hacks

Where a mature international solution exists, we use it instead of inventing our own protocol or platform. Before writing code, we study how the problem is already solved in the industry.

Transferability

The solution is loosely coupled and documented: it can be handed over between teams and contractors without rewriting. You are not tied to us.

AI compatibility

n8n in the AI stack

The agent decides WHAT, n8n executes HOW

LLM agents select actions (non-deterministic reasoning); n8n transforms them into reproducible steps: permitted actions, retries, idempotency, secrets, least-privilege—agents never blindly call production APIs.

AI Agent based on LangChain

The AI Agent node (Tools, ReAct, SQL, Conversational, and others) builds agent workflows with tools and memory directly on the canvas, without a separate framework.

17 LLM providers, including Anthropic Claude

OpenAI, Anthropic, Google Gemini/Vertex, Mistral, Bedrock, plus local models via Ollama — swap the model without rewriting the workflow.

RAG on 12 vector stores

Pinecone, Qdrant, PGVector, Weaviate, Redis, and more connect as nodes — search over enterprise data is built into the workflow.

AI as a process step, not a separate product

The LLM call sits between the trigger and the action in the system, so AI immediately affects a real operation instead of staying a demo.

Local models and privacy

With a self-hosted deployment, AI calls can be kept inside your perimeter (Ollama, local inference), so data does not go to an external provider.

News

What's new in n8n

All news

2026-08-25
AI Assistant on self-hosted simplified to a single command in 2.35
Previously, running AI Assistant self-hosted required manually enabling the instance-ai module and configuring the sandbox, model, and web search through environment variables. In 2.35, the module is enabled by default, installation takes one command, and the model provider (Anthropic, OpenAI, OpenRouter, or any OpenAI-compatible endpoint, including local ones), sandbox, and web search are configured directly in the UI.

n8n lets each instance choose its model, including a local OpenAI-compatible endpoint, without requiring the vendor's cloud; it does not provide centralized routing or quotas across multiple instances like a dedicated gateway. LLM Gateway →
2026-08-20
A security update fixed 9 vulnerabilities, including RCE via $fromAI and sandbox escape
Nine security advisories in one release: 6 High (including an Expression Sandbox Escape via a prototype leak in $fromAI leading to Host RCE — GHSA-9x83-43r8-5hwc; RCE in the Git node; credential leaks from Strapi, SeaTable, and Mailcheck nodes; SSRF and local file reads through Gmail/Brevo) and 3 Medium (including query injection in Elasticsearch/Firestore nodes). Fixed in v1.123.73, v2.35.4 (stable), and v2.36.2 (beta); Cloud was patched automatically, while self-hosted instances require manual updates.

The vulnerability is not in Elasticsearch itself, but in the n8n connector node for it (query injection): the risk lies in the integration layer, not the search engine. Elasticsearch →
2026-06-08
OpenTelemetry tracing for workflow and node execution
Each workflow run creates a root span with status, duration, and error, plus an optional child span for each node. Enabled with two env variables, exported to any OTLP collector (Jaeger, Tempo, Datadog, Honeycomb, Grafana Cloud) with no vendor lock-in.

Kafka currently has no native OTLP tracing on the broker side; it is usually instrumented through client libraries or Kafka Connect. n8n provides tracing out of the box with two environment variables. Apache Kafka →
2026-05-19
Managed connection to MCP servers without manual client setup
n8n 2.22.0: the agent connects to some external MCP servers (Apify, Linear, Notion, and others from the official MCP registry) by selecting them from the node panel and logging in, without manually creating an MCP Client node and credential.

Lowers the entry barrier to the MCP protocol to selecting a connector from a list, removing the manual client+credential setup that was previously required even for standard MCP servers MCP →