# n8n

> n8n is a workflow automation platform that connects apps, APIs, and AI models on a visual node canvas, with full code access (JavaScript and Python) when the visual layer runs out. It is source-available under a fair-code license, so it can be self-hosted for free, and its cloud pricing charges per workflow execution rather than per step, which makes complex multi-step automations dramatically cheaper than task-priced rivals.

- Category: GTM Engineering (https://saastracker.org/categories/gtm-engineering)
- Website: https://n8n.io
- Starting price: About $24/mo (Cloud Starter, 2,500 executions; roughly $20 on annual billing)
- Free plan: Self-hosted Community edition is free and unmetered on your own infrastructure
- Free trial: Cloud trial, no card required
- Founded: 2019, HQ: Berlin, Germany, Ownership: Venture-backed, founder-led
- Profile last reviewed: 2026-08-22
- Canonical profile: https://saastracker.org/products/n8n

## Overview

Every automation platform draws a line between what the visual builder can do and where you need an engineer. n8n's bet is to put that line as far out as possible: the canvas handles branching, loops, merges, and error paths natively, and when you do hit the edge, a code node runs real JavaScript or Python inside the workflow instead of forcing you out to a serverless function. That makes it the automation tool technical teams graduate to when Zapier's linear Zaps or Make's operation counts start fighting the work.

Two structural decisions define the economics. First, fair-code licensing: the source is public and the Community edition self-hosts for free, so the ceiling on cost is your own server. Second, per-execution pricing in the cloud: one workflow run costs one execution whether it has three steps or three hundred. A fifty-step enrichment waterfall that would consume fifty tasks on Zapier or fifty operations on Make bills as a single execution, which is exactly why n8n has become the default engine for Clay-style DIY pipelines, where workflows are long by nature.

The company, founded in Berlin in 2019 by Jan Oberhauser, spent its first years as a beloved developer tool and then compounded hard when AI agents arrived: LangChain-based agent nodes made n8n one of the easiest places to give a language model tools, memory, and guardrails inside a production workflow. Growth followed, a $60M Series B led by Highland Europe in early 2025, and a reported $180M round later that year at a valuation around $2.5B, making n8n one of Europe's fastest-appreciating software companies while the hosted starter plan still costs about what two streaming subscriptions do.

## How it works

1. You build workflows on a node canvas: a trigger node starts the run (a schedule, a webhook, an app event, a chat message, or a manual click), and each subsequent node transforms data or calls a service. Over 400 built-in integrations cover the common SaaS surface, a generic HTTP Request node covers every other API, and code nodes run JavaScript or Python mid-flow. Data passes between nodes as JSON items you can inspect at every step, which makes debugging concrete: you see exactly what each node received and produced on the last run.

2. Control flow is native rather than bolted on: IF and Switch nodes branch, Merge joins parallel paths, Loop nodes iterate over batches, and error workflows catch failures and route them to alerts or retries. Sub-workflows let you compose one tested pipeline inside another, so a 'find email and verify' block can be reused across every campaign workflow that needs it.

3. The AI layer is built on LangChain: agent nodes wrap a model (OpenAI, Anthropic, Google, or a local one) with tools, memory, and structured output parsing, so an agent can decide which of your workflow branches to take, qualify a lead in one node, or run a full retrieval pipeline against a vector store. Because the agent lives inside a normal workflow, everything around it, rate limits, retries, logging, human approval steps, is ordinary n8n.

4. You run it where you want: n8n Cloud is the hosted option with per-execution pricing, or you self-host the Community edition free via Docker or npm, keeping every byte of data on your own infrastructure. Paid self-hosted tiers add collaboration, SSO, and environments; an embed license lets SaaS products ship n8n as their own integration layer.

## Best for

Technical or semi-technical teams building long, multi-step automations, enrichment waterfalls, AI agents, and internal integrations, who want per-execution economics in the cloud or full data control self-hosted, and who are comfortable reading JSON when something breaks.

## Not the right fit for

- Non-technical users who want the automation tool to disappear; Zapier's guided setup and enormous app catalog make the first Zap much easier than the first n8n workflow.
- Teams that need a specific long-tail SaaS connector today; n8n's 400+ integrations cover far less surface than Zapier's thousands, and the fallback is the HTTP node, which assumes API literacy.
- Anyone for whom 'self-hosted' means 'someone else's problem'; the free Community edition is genuinely free only if maintaining a server, upgrades, and backups costs you nothing.
- Strict open-source purists; fair-code is source-available with commercial restrictions, not an OSI-approved license, which matters to some procurement and ideology.
- One-step automations at low volume; if the job is 'new form entry, send Slack message', task-priced tools are simpler and their free tiers are enough.

## Features

### Workflow building

A node canvas that keeps going where linear builders stop.

- **Visual node editor**: Workflows are graphs, not lines: branch, merge, loop, and run paths in parallel on one canvas, with per-node input and output data visible from the last execution.
- **Code nodes (JavaScript and Python)**: Run real code mid-workflow with access to the item stream, npm-style helpers, and expressions, removing the ceiling visual-only builders hit.
- **400+ native integrations**: Built-in nodes for the common SaaS surface: Google Sheets, Slack, HubSpot, Notion, Airtable, Postgres, Telegram, and hundreds more, each exposing operations rather than just triggers.
- **HTTP Request node**: A first-class generic API client with authentication presets, pagination handling, and cURL import, which effectively makes every API an integration.
- **Webhook triggers**: Instant inbound webhooks with test and production URLs, so external systems can start workflows in real time rather than on a polling delay.
- **Sub-workflows**: Call one workflow from another with typed inputs and outputs, letting teams build a library of tested blocks (enrich, verify, notify) reused across pipelines.
- **Error workflows and retries**: Failures route to dedicated error workflows with the full failed payload; nodes retry with backoff, and runs can resume from the failed step instead of restarting.
- **Templates library**: Thousands of community workflow templates import in one click, which is how most teams learn the idioms: copy a working pipeline, then bend it.

### AI and agents

LangChain built into the canvas, so agents ship inside real workflows.

- **AI Agent nodes**: Wrap any major model (OpenAI, Anthropic, Google, or self-hosted via Ollama) with tools, memory, and a system prompt; the agent's tool calls are other n8n nodes.
- **Model-agnostic LLM chains**: Prompt, summarize, classify, and extract with structured output parsers that validate the model's JSON before it touches the rest of the pipeline.
- **Vector store and RAG nodes**: Load, embed, and query documents against Pinecone, Qdrant, Supabase, and in-memory stores for retrieval-augmented workflows without external glue code.
- **Chat triggers**: A hosted chat UI or embedded widget can front any workflow, turning an automation into an internal copilot with the workflow as its brain.
- **Human-in-the-loop steps**: Approval nodes pause a run until someone confirms in Slack, email, or a form, the pattern that makes AI-written outreach safe to ship.

### Deployment and economics

The two decisions that define n8n: run it anywhere, pay per run.

- **Per-execution pricing**: A workflow run is one execution regardless of step count, so fifty-node waterfalls cost the same as three-node notifications, the inverse of task and operation pricing.
- **Free self-hosted Community edition**: The full core engine runs on your own server via Docker or npm at no license cost under the fair-code Sustainable Use License.
- **n8n Cloud**: The hosted service handles upgrades, scaling, and uptime, with plans from about $24 a month (roughly $20 on annual billing) for 2,500 executions.
- **Unlimited workflows and steps**: Plans meter executions, not the number of workflows, users on the canvas, or nodes per workflow, so building more does not cost more until it runs more.
- **Embed license**: SaaS companies can white-label the n8n editor inside their own product as the customer-facing integration layer.

### Collaboration and operations

The production apparatus around the canvas.

- **Projects and RBAC**: Workflows organize into projects with role-based access, separating a marketing team's pipelines from finance's on one instance.
- **Environments and Git sync**: Source-control workflows and promote them dev to production on higher tiers, treating automations as deployable code rather than shared drafts.
- **Execution history and logs**: Every run is stored with full per-node data for inspection and replay, which turns 'why did row 412 fail' from archaeology into a click.
- **Credentials vault**: API keys and OAuth connections are stored encrypted once and referenced by nodes, with sharing controls so keys do not spread through the team.
- **SSO, SAML, and LDAP**: Enterprise identity plumbing on paid business tiers, alongside audit logging and external secrets integration.

## Use cases

- **Founder replacing a Clay subscription**: Runs a domain-to-verified-email enrichment waterfall (search, qualify with GPT, find email, verify) and cannot justify $350 a month for a managed table when the pipeline is five API calls. Outcome: The whole waterfall is one n8n workflow triggered from a Google Sheet; at per-execution pricing a 1,000-domain run costs a fraction of the equivalent task-priced bill, and the $24 hosted plan replaces the Clay line item. This exact stack is written up in SaaSTracker's enrichment waterfall playbook.
- **RevOps lead at a 30-person startup**: CRM, billing, and support systems disagree about customers; the integration budget is one person's spare afternoons, and every new SaaS tool adds another sync to maintain. Outcome: n8n becomes the integration bus: webhook-triggered workflows keep HubSpot, Stripe, and the warehouse consistent, error workflows page Slack when a sync breaks, and sub-workflows keep the common lookups in one tested place.
- **Agency building AI agents for clients**: Clients want lead qualification bots and internal copilots, but every prototype built on raw API calls collapses under retries, logging, and approval steps the demo never had. Outcome: Agent nodes supply the model wiring while the surrounding workflow supplies production hygiene: rate limits, human approval before anything sends, and execution logs the client can audit. The agency self-hosts per client for data isolation at zero license cost.
- **Privacy-sensitive European SMB**: Wants automation across internal tools but cannot send customer data through a US cloud automation vendor, and the compliance review kills every hosted option. Outcome: Self-hosted Community edition keeps every payload on the company's own EU server; the workflow catalog and HTTP node cover the internal APIs, and the only recurring cost is the VM.

## Pricing

Two paths. Cloud: subscription tiers metered by workflow executions per month (a run is one execution regardless of step count), with unlimited workflows and steps on every tier. Self-hosted: the Community edition is free under the fair-code Sustainable Use License, with paid Business and Enterprise self-hosted tiers adding collaboration, SSO, environments, and support. Prices are quoted in euros; dollar figures shift slightly with exchange rates.

- **Community (self-hosted)**: Free your own server. Full core engine: canvas, 400+ integrations, code nodes, AI agent nodes; Unlimited workflows and executions, bounded only by your hardware; Fair-code licensed; you handle hosting, upgrades, and backups.
- **Starter (Cloud)**: ~€20 to €24 per month, annual vs monthly billing. 2,500 executions/month, unlimited workflows and steps; Hosted, upgraded, and backed up by n8n; The plan the DIY enrichment stack actually needs.
- **Pro (Cloud)**: ~€50 to €60 per month, annual vs monthly billing. 10,000 executions/month with higher concurrency; Projects, longer execution log retention, admin roles; Scales via execution add-on packs.
- **Business / Enterprise**: Custom cloud or self-hosted. SSO/SAML, LDAP, environments with Git sync, external secrets; Audit logs, dedicated support, embed licensing available.

Billing notes:

- Per-execution metering is the headline economic fact: a 50-step workflow run bills as 1 execution, where task-priced tools would bill 50. The longer your workflows, the bigger the gap.
- Every cloud tier includes unlimited workflows, steps, and (on Starter) the full node catalog including AI nodes; tiers differ on executions, concurrency, and ops features.
- Prices are set in euros, so dollar costs float with the exchange rate; annual billing saves roughly 17%.
- Self-hosting is free in license terms but not in practice: budget the server (a small VPS suffices for modest loads) and someone's time for upgrades.
- The Sustainable Use License permits internal business use freely; reselling n8n itself as a hosted service requires the commercial embed license.

Value assessment: For long workflows, nothing in the category touches n8n's economics: the per-execution meter makes a 40-step enrichment waterfall cost the same as a 2-step alert, and the same pipeline on task or operation pricing can cost ten to fifty times more at identical volume. Against that, the true cost is skill: n8n assumes you can read JSON, debug an HTTP call, and occasionally write a line of code. Teams with that skill get Zapier-class capability at a fraction of the spend, plus a free self-hosted exit ramp that caps vendor risk permanently. Teams without it will burn the savings in time.

## Strengths

- Per-execution pricing inverts the category's cost model; long, complex workflows, exactly the kind GTM engineering produces, are where it is cheapest relative to rivals.
- Free self-hosted Community edition provides a permanent exit ramp and full data sovereignty, a structural negotiating position no closed rival offers.
- Code nodes and the HTTP Request node remove the capability ceiling; if an API exists, n8n can drive it without leaving the canvas.
- LangChain-based agent nodes made it one of the fastest paths from 'AI idea' to 'AI in production', with approvals, retries, and logging around the model.
- Graph-shaped workflows (branch, merge, loop, error paths) model real processes that linear builders force into workarounds.
- Execution history with full per-node payloads makes debugging concrete rather than forensic.

## Limitations

- The learning curve is real: items, expressions, and JSON structure confront you in week one, and non-technical users often stall where Zapier would have carried them.
- The integration catalog (400+) is an order of magnitude smaller than Zapier's; long-tail SaaS tools frequently mean building against the raw API via the HTTP node.
- Fair-code is not open source by OSI definition, which complicates some corporate policies and community expectations, and the license debate resurfaces periodically.
- Cloud Starter's 2,500 executions can pinch for high-frequency polling triggers; trigger design (webhooks over polling) becomes a cost skill.
- Self-hosting shifts upgrade, backup, and security burden onto you, and major version upgrades occasionally require migration attention.
- Euro-denominated pricing makes US budgeting slightly fuzzy, and the exact dollar figure moves between billing cycles.

## Comparisons

- **n8n vs Zapier**: Zapier is the easiest way to connect two apps; n8n is the cheapest way to connect twenty steps. Zapier wins on catalog breadth (thousands of apps vs hundreds), polish, and time-to-first-automation for non-technical users. n8n wins the moment workflows get long or logical: per-execution pricing versus per-task pricing can be an order-of-magnitude cost gap on multi-step pipelines, code nodes remove the ceiling, and self-hosting removes the vendor. Ops teams standardize on Zapier; GTM engineers standardize on n8n.
- **n8n vs Make**: The two power builders. Make gives you a visual scenario canvas with the category's cheapest entry pricing, but meters every operation, so long waterfalls still accumulate cost per step. n8n meters the run, not the steps, adds real code nodes and self-hosting, and has pulled decisively ahead on AI agent tooling. Make suits visually-minded builders who want power without code; n8n suits builders who will eventually want the code, and whose workflows are long enough for per-execution pricing to pay.

## Implementation

- Setup time: Cloud: minutes to sign up, and a first webhook-to-Slack workflow the same hour. Self-hosted: a Docker one-liner for a test instance; a hardened production deployment with backups and HTTPS is a half-day for someone who has run containers before.
- Learning curve: Steeper than task-based tools: expect a few days to internalize items, expressions, and node data mapping, typically via imported templates. Engineers acclimate almost immediately; operators without API experience need the templates and patience.
- Onboarding: Self-serve with documentation, an active community forum, and a large template library; paid business tiers add dedicated support. No mandatory onboarding calls at self-serve tiers.
- Migration: Migrating from Zapier or Make means rebuilding workflows by hand, there is no importer, but teams usually report the rebuild is faster than the original build because the logic is already designed. Workflows export as JSON, so moving between n8n cloud and self-hosted is trivial in both directions.

## Platform, API & security

- Platforms: Web app (cloud), Self-hosted (Docker, npm), Embedded (commercial license), REST API
- API: A public REST API manages workflows, executions, and credentials programmatically; workflows themselves are portable JSON. Webhook and HTTP nodes make n8n both an API consumer and an API producer.
- Compliance: SOC 2, GDPR
- Data residency: Cloud hosting with EU-based options; self-hosting places data wherever you run it, which is the definitive residency answer for regulated teams.
- SSO: SAML SSO and LDAP on Business and Enterprise tiers.
- Security notes: The self-hosted option is the security story: payloads, credentials, and logs can remain entirely on your infrastructure. Cloud tenants get encrypted credential storage and standard SaaS controls.

## Support

- Channels: Community forum, Email support (cloud plans), Dedicated support (Business and Enterprise)
- Documentation: Extensive documentation covering every node, hosting, and scaling, plus thousands of importable community templates that function as executable documentation.
- Community: One of the category's strongest: a very active forum, a large Discord presence, and a template ecosystem that materially shortens the learning curve.

## Company

- Founded: 2019
- Founders: Jan Oberhauser
- Headquarters: Berlin, Germany
- Ownership: Venture-backed, founder-led
- Employees: Several hundred (2026)
- Funding: Roughly $250M raised, including a $60M Series B led by Highland Europe (early 2025) and a reported $180M round later in 2025 at a valuation around $2.5B; earlier backers include Sequoia and Felicis.

Funding history:

- Seed (2020): $1.5M. Sequoia's first seed investment in a German company
- Series A (2021): $12M. Led by Felicis
- Series B (2025): $60M. Led by Highland Europe as AI-agent adoption accelerated
- Series C (2025): $180M (reported). Reported at a ~$2.5B valuation led by Accel

Timeline:

- 2019: Jan Oberhauser open-sources n8n in Berlin under a fair-code license; 'nodemation' becomes n8n.
- 2020: Sequoia leads a $1.5M seed, its first in a German startup; the community and node catalog compound.
- 2021: $12M Series A led by Felicis; n8n Cloud launches so teams can run it without self-hosting.
- 2023: LangChain-based AI nodes land, turning n8n into one of the earliest production homes for LLM workflows and agents.
- 2025: $60M Series B (Highland Europe), then a reported $180M round at a ~$2.5B valuation as n8n becomes the default engine of the GTM engineering stack.
- 2026: Per-execution pricing, agent tooling, and the template ecosystem make n8n the reference answer to 'do I really need to pay per task?'

## Integrations

Google Sheets, Slack, HubSpot, Salesforce, Notion, Airtable, OpenAI, Anthropic, Telegram, Discord, Postgres, MySQL, Stripe, Gmail, Webhooks / any REST API (HTTP node)

## FAQ

### What is n8n?

n8n is a workflow automation platform that connects apps, APIs, and AI models on a visual node canvas, with JavaScript and Python code nodes for anything the visual layer cannot express. It is source-available under a fair-code license, self-hostable for free, and its cloud plans charge per workflow execution rather than per step.

### How much does n8n cost?

The self-hosted Community edition is free on your own server. n8n Cloud starts at about €20 a month on annual billing (roughly €24 monthly, around $24) for 2,500 executions with unlimited workflows and steps; the Pro tier at about €50 covers 10,000 executions, and Business and Enterprise tiers add SSO, environments, and support at custom pricing.

### What does per-execution pricing mean in practice?

One run of a workflow counts as one execution no matter how many steps it contains. A 50-step enrichment waterfall processed once costs one execution on n8n; the same pipeline on task-priced or operation-priced tools bills every step, which is why long workflows are where n8n's economics dominate.

### Is n8n open source?

It is source-available under the fair-code Sustainable Use License: you can read the code, self-host freely for internal business use, and modify it, but you cannot resell n8n itself as a hosted service without a commercial license. That is not OSI-approved open source, which matters to some buyers and not at all to most.

### Is n8n hard to learn?

Harder than Zapier, easier than writing the scripts yourself. Expect a few days to get comfortable with items, expressions, and data mapping between nodes; most people learn by importing one of the thousands of community templates and modifying it. Engineers are typically productive on day one.

### Can n8n build AI agents?

Yes, and it is one of the platform's biggest draws: LangChain-based agent nodes wrap OpenAI, Anthropic, Google, or local models with tools, memory, and structured outputs, while the surrounding workflow provides retries, rate limits, logging, and human-approval steps. The agent's tools are simply other n8n nodes.

### How many integrations does n8n have?

Over 400 native integrations, plus a first-class HTTP Request node that can call any REST API with authentication presets and cURL import. The native catalog is much smaller than Zapier's thousands of apps, but the HTTP node means coverage is limited by APIs, not by the catalog.

### Should I self-host n8n or use n8n Cloud?

Self-host if data control matters (regulated data, EU-only processing) or if you already run infrastructure and want zero license cost. Use Cloud if you would rather pay about $24 a month than maintain a server; you can switch later, since workflows export as portable JSON in both directions.

### Is n8n cheaper than Zapier or Make?

For multi-step workflows, usually dramatically: n8n bills the run while Zapier bills each task and Make bills each operation, so a long waterfall can cost ten to fifty times less on n8n at the same volume. For short two-step automations at low volume the difference shrinks, and free tiers elsewhere may be simpler.

### Who is behind n8n?

n8n was founded in Berlin in 2019 by Jan Oberhauser. Backers include Sequoia, Felicis, and Highland Europe, with roughly $250M raised including a reported $180M round in late 2025 at a valuation around $2.5B, making it one of Europe's fastest-growing software companies.

## Editorial verdict

n8n is what the automation category looks like when it is priced and licensed in the buyer's favor: pay per run instead of per step, read the source, and leave for your own server whenever you like. For the GTM engineer building long enrichment waterfalls and AI-agent workflows, those three facts compound into the category's best deal, and the LangChain-native agent tooling has made it the default engine of the DIY stack. The honest cost is skill: it asks more of you in week one than Zapier ever will, and its integration catalog still leans on the HTTP node for the long tail. Teams with a technical operator should default here; teams without one should pay the Zapier tax knowingly.

## SaaSTracker awards

- Momentum (GTM Engineering, Summer 2026): "Per-execution pricing, native AI agent nodes, and a free self-hosted edition made it the engine of the DIY enrichment stack, and a reported $2.5B valuation says the market noticed."

---

Source: SaaSTracker (https://saastracker.org), an independent editorial project. This profile is compiled from public information, carries no peer reviews or paid placement, and was last reviewed 2026-08-22. Awards are judged on published criteria: https://saastracker.org/methodology