Is n8n actually free?

The Community Edition (self-hosted) is fair-code licensed and free with unlimited executions. You provide the infrastructure — typically $5-20/mo for a small DigitalOcean, Hetzner, or AWS Lightsail instance, plus your own time to maintain it. The cloud-hosted plans start at €24/mo (Starter, 2,500 executions) and scale to €60/mo (Pro, 10,000) and €800/mo (Business, 40,000). Self-host wins on cost above ~10k executions/mo if you have any DevOps capacity; cloud wins on convenience below that.

How does n8n actually compare to Zapier on cost?

The structural difference is execution counting. n8n charges per workflow execution (one full workflow run). Zapier charges per task (every step is a task). A 10-step workflow that runs 1,000 times = 1,000 n8n executions but 10,000 Zapier tasks. At Zapier's $0.025-0.05/task pricing, that's $250-500/mo on Zapier vs ~$24-60/mo on n8n cloud or ~$10/mo on n8n self-host. The cost gap compounds with workflow complexity, which is why teams running multi-step automations migrate quickly once they do the math.

Do I really need a developer to run self-hosted n8n?

You need someone comfortable with Docker basics or basic cloud server management — that's a 1-3 hour setup, not a multi-week project. Once running, the visual workflow editor is accessible to non-developers. The two real DevOps tasks are: (1) initial install on a VPS with HTTPS and a reverse proxy, (2) ongoing updates and backups (~30 min/quarter). If your team has zero DevOps capacity, n8n cloud (€24/mo Starter) skips this entirely and you still get most of the cost advantage vs Zapier.

What about polling triggers eating my execution budget?

This is the most common pricing surprise on n8n cloud. A polling trigger checking every 5 minutes runs 12 times/hour × 24 × 30 = 8,640 executions/month — that single workflow exhausts the Starter plan (2,500/mo) by day nine and burns most of the Pro plan (10,000/mo). Solutions: (1) use webhooks instead of polling wherever possible, (2) reduce poll frequency to hourly or daily for non-time-critical flows, (3) run polling-heavy workflows on self-hosted (no execution cap), (4) consolidate multiple polls into a single scheduled workflow that branches.

How does n8n compare to Make (formerly Integromat)?

Both are visual workflow automation tools. Make is the middle ground — more polished UX than n8n, more powerful than Zapier, ~1,500 integrations. n8n has fewer native integrations (~1,000) but its HTTP node + custom code support means it can talk to anything with an API. Make charges per operation (similar trap to Zapier's per-task model); n8n charges per execution. For SMBs that prefer a polished SaaS UX, Make is reasonable. For technical teams or anyone planning workflows that scale, n8n's pricing model is structurally better.

What integrations does n8n actually support?

About 1,000 native integration nodes covering the major SaaS tools (Slack, Gmail, Salesforce, HubSpot, GitHub, Notion, Stripe, Shopify, Postgres, MySQL, MongoDB, etc.) plus generic HTTP request, webhook, and code execution nodes that connect to anything with an API. The catalog is smaller than Zapier (~6,000) and Make (~1,500), but the HTTP + code nodes close most of the gap for teams comfortable reading API docs. The integration depth varies — some integrations are surface-level OAuth wrappers, others have dozens of well-modeled actions.

Is n8n suitable for compliance-heavy industries?

Yes — self-hosting is one of the main reasons regulated industries choose n8n. Your data stays on infrastructure you control, which simplifies SOC2, HIPAA, GDPR data residency, and ITAR-style requirements. The fair-code license permits self-host for internal use without restriction. n8n cloud is also SOC2 Type 2 certified for the cloud-hosted tier. If your buyer requires data sovereignty, n8n is structurally better positioned than any per-task SaaS tool that must process your data on their infrastructure.

How well does n8n handle AI agent workflows in 2026?

This is where n8n has become a serious player in the last 12 months. The AI agent layer ships first-class support for multi-step agent workflows, native nodes for OpenAI / Anthropic / Google Gemini, vector store integrations (Pinecone, Qdrant, Weaviate), and chain composition that's closer to LangChain than to Zapier's natural-language workflow builder. For teams building agent-style automations (research workflows, content pipelines, support triage), n8n is competitive with purpose-built agent frameworks while staying inside a visual workflow editor.

Can n8n replace Workato or Zapier outright?

For most use cases, yes. The migration overhead depends on workflow complexity and integration coverage. Teams running 30-50 active Zapier automations typically take 2-6 weeks for a clean migration if the integrations are all in n8n's catalog. If you depend on long-tail integrations (the 5,000+ niche apps Zapier supports that n8n doesn't), the migration includes building those connections via HTTP nodes. Cost savings usually pay back within the first quarter — often the first month at scale.

What's the catch — what does n8n not do well?

Three real weaknesses. (1) Documentation can be thin or out of date for newer features; community forum and Discord fill the gap but expect to dig. (2) Error messages are sometimes vague ("Problem executing workflow") and require manually inspecting the canvas to find the failed node — frustrating during initial debugging. (3) The learning curve is real for non-technical users — concepts like webhook authentication, JSON traversal, and execution branching assume some technical comfort. If your automations are owned by a non-technical marketing or ops team without engineering support, Zapier or Make's gentler curves win.

StackSwap review · published 2026-05-06

n8n Review (2026): Honest Take After Running It at Scale

n8n is the open-source workflow automation platform we recommend for engineering-led teams, anyone hitting Zapier's per-task pricing wall, and operators building AI agent workflows. Self-host is free with unlimited executions; cloud starts at €24/mo. Here's the operator's take after running it for our own internal automations, plus how it really compares to Zapier, Make, Workato, Tray, and Pipedream — and the polling-trigger trap nobody warns you about.

Editorial score: 4.4/5 — strong fit for engineering-led ops teams, compliance-heavy use cases, and AI agent builders. Soft pass for non-technical teams without engineering support.

The TL;DR

If you're running multi-step workflows at any volume on Zapier or Make and the monthly bill is north of $200, n8n almost certainly cuts your cost by 5-10x — sometimes more — without losing meaningful capability. The catch is that n8n assumes more technical comfort than its competitors: comfortable with webhook auth, JSON paths, and the occasional Docker command if you self-host. If your team has that comfort, it's a no-brainer. If your team is non-technical operators with no engineering support, stick with Zapier or Make and pay the convenience tax.

The 2026 angle: n8n's AI agent layer (native nodes for OpenAI / Anthropic / Gemini, vector stores, chain composition) makes it competitive with purpose-built agent frameworks while staying inside a visual workflow editor. For teams building agent-style automations — research workflows, content pipelines, support triage — this is the most interesting development in the category in two years.

The pricing math vs Zapier (the structural difference)

The single most important thing to understand about n8n vs Zapier: execution counting. Zapier counts every step in a workflow as a billable task. n8n counts a full workflow run as one execution. A 10-step workflow that runs 1,000 times = 1,000 n8n executions, but 10,000 Zapier tasks. At Zapier's pricing this is a 10x differential before you even compare list prices.

Workflow scale	Zapier (per task)	n8n cloud (per execution)	n8n self-host
2,000 simple workflow runs/mo (5 steps each)	$50-100/mo (10k tasks)	€24/mo (Starter, 2.5k execs)	~$5/mo (small VPS)
10,000 workflow runs/mo (8 steps each)	$300-600/mo (80k tasks)	€60/mo (Pro, 10k execs)	~$10-20/mo (small VPS)
50,000 workflow runs/mo (10 steps each)	$1,500-3,000/mo (500k tasks)	€800/mo (Business, 40k) or contact sales	~$50-100/mo (dedicated)
500,000+ workflow runs/mo (scale)	$5,000-15,000/mo enterprise	Custom enterprise pricing	~$150-300/mo (dedicated cluster)

Zapier task counts assume 5-10 steps per workflow at average step density. n8n cloud pricing reflects published tiers as of mid-2026 (Starter €24, Pro €60, Business €800). Self-host figures assume small DigitalOcean / Hetzner / AWS Lightsail instances sized to load, plus Postgres for production. Confirm current pricing on each vendor's site.

Real stacked pricing — what teams actually pay

The headline pricing tells one story; the realistic functional cost is what matters. Here's how the math actually plays out for the stacks we see most often:

Setup	Plan	Add-ons / infra	Real monthly cost
Solo founder, 1-3 simple workflows	Self-host (free Community Edition)	$5/mo VPS (DigitalOcean, Hetzner)	~$5/mo
Solo founder who wants zero DevOps	Cloud Starter (€24/mo, 2,500 executions)	None for low-volume	~€24/mo (~$26)
Lean ops team, 10-30 workflows, moderate volume	Cloud Pro (€60/mo, 10,000 executions)	Watch polling triggers; consider self-host	~€60/mo (~$66)
Engineering-led team, 50+ workflows or polling-heavy	Self-host on small dedicated server	$20-50/mo (DigitalOcean droplet, Postgres, backups)	~$30-60/mo
Mid-market team, high volume, no DevOps capacity	Cloud Business (€800/mo, 40,000 executions)	SSO, RBAC, log streaming included	~€800/mo (~$880)
Scale (500k+ executions/mo, multiple environments)	Self-host dedicated + monitoring	$100-200/mo infra + DevOps time	~$150-300/mo

The polling-trigger trap (read this before you pick a plan)

This is the single most common pricing surprise on n8n cloud and the docs don't warn you well enough. A polling trigger that checks every 5 minutes runs:

12 polls/hour × 24 hrs × 30 days = 8,640 executions/month

That single workflow exhausts the Starter plan (2,500/mo) by day nine and consumes most of the Pro plan (10,000/mo) without any other workflows running. New cloud customers blame the platform pricing when the real culprit is workflow design.

The fix is structural:

Use webhooks instead of polling wherever possible. A webhook fires once when an event happens, not every 5 minutes.
Drop poll frequency. If hourly polling works, that's 720 executions/month — manageable on Starter even with multiple flows.
Move polling-heavy workflows to self-host. No execution cap, so polling is free. The simplest fix once you cross the cost threshold.
Consolidate multiple polls into one scheduled workflow that branches. Three separate 5-min polls = 25,920 executions/mo. One scheduled flow that handles all three branches = 8,640 executions/mo (1/3 the cost).

Where n8n shines

Per-execution pricing instead of per-task. The structural cost advantage versus Zapier and Make. A 10-step workflow that runs 1,000 times = 1,000 n8n executions but 10,000 Zapier tasks or Make operations. At scale this is the difference between a $24-60/mo cloud bill and a $300-500/mo Zapier bill for equivalent work. The pricing model alone justifies the migration for any team running multi-step automations at volume.
Self-host eliminates the per-execution ceiling entirely. Community Edition is free and runs unlimited executions on your own server. A $5-20/mo VPS is enough for most SMB workloads; a $50-100/mo dedicated server handles mid-market scale. No surprise overage bills when a webhook trigger goes runaway and burns 50,000 executions in an hour. Self-hosting also means data sovereignty, which matters in compliance-sensitive industries.
Code-extensible nodes for engineering teams. JavaScript code nodes let you write custom logic inline — no need to hand off to a dedicated backend service for one-off transformations. Function nodes can call external libraries (with self-host), parse complex payloads, and implement business logic that would otherwise require a custom Lambda or Cloud Function. For engineering teams, this is the difference between a workflow builder and a real automation runtime.
Strong API + webhook handling. Better suited to complex pipelines than Zapier's task-shaped abstraction. Webhook receivers, conditional branching, error handling, retry policies, and queue-mode execution are all first-class. The HTTP request node handles any REST API; the OAuth helpers cover common auth flows. Engineering teams treat n8n as a lightweight orchestration runtime, not a no-code toy.
AI agent layer is competitive with purpose-built frameworks. Native nodes for OpenAI, Anthropic, Google Gemini; vector store integrations (Pinecone, Qdrant, Weaviate, Supabase); chain composition that supports tool-calling agents and multi-step reasoning. For teams building agent-style automations (research workflows, content pipelines, support triage), n8n is competitive with LangChain or LlamaIndex while staying inside a visual editor — meaningful productivity advantage if your team is more comfortable with workflows than code.
Active open-source community. Fair-code license keeps the project sustainable while preserving free self-hosting for internal use. Active GitHub (45k+ stars), Discord, and community forum. Community-built nodes ship for niche integrations that the core team hasn't prioritized. The community is part of the product — you'll often find a working pattern in the forum before the docs cover it.
Lead-funded, no enshittification pressure (yet). n8n raised $55M Series B in 2024, so the company has runway and isn't pressured to monetize self-hosters. The fair-code license includes some restrictions on running n8n as a paid SaaS for others, but doesn't restrict internal use. This is a real consideration vs other open-source tools that rug-pull their license once they need to monetize.

Where it breaks (the honest list)

Polling triggers can blow through cloud execution budgets. A polling trigger that checks every 5 minutes burns 8,640 executions/month — that single workflow exhausts the Starter plan (2,500/mo) by day nine and consumes most of the Pro plan (10,000/mo). Operators new to n8n cloud get blindsided by this and assume the platform is overpriced. The fix is structural: use webhooks instead of polling, drop poll frequency to hourly, or move polling-heavy workflows to self-host. But the trap is real and the docs don't surface it well.
Documentation is uneven. Core node docs are reasonable but newer features (especially AI agent nodes) lag behind the actual implementation. The community forum and Discord are often more current than the official docs, which means you spend more time searching threads than reading reference material. Improving but not yet at Zapier's level of polish.
Error messages are sometimes opaque. When a workflow fails, you often get a generic "Problem executing workflow" message that requires opening the canvas, finding the failed node, and inspecting its output to debug. For complex workflows with 20+ nodes, this gets tedious. Better error surfacing is on the roadmap but not shipped as of mid-2026. Expect to spend more debugging time than on Zapier or Make.
Self-host requires real DevOps discipline. Production self-hosting means: HTTPS via reverse proxy, Postgres for the main database (not the default SQLite), Redis for queue mode at scale, log aggregation, backups, version upgrades. None of this is hard, but if your team has no DevOps capacity, the "free" self-hosted tier costs more in time than the cloud plans cost in money. Don't self-host n8n as your first DevOps project unless you have someone who can own it.
Integration catalog is smaller than Zapier or Make. ~1,000 native integrations vs Zapier's 6,000+ and Make's 1,500. The HTTP node closes most of the gap for teams comfortable with API docs, but if you depend on a long-tail SaaS tool with no native node, you're building the integration yourself. For mainstream B2B tools (Salesforce, HubSpot, Slack, Stripe, etc.) the catalog is fine. For niche industry-specific apps, check the catalog before committing.
UX is targeted at technical users. The visual editor is approachable but the surrounding context (webhook authentication, JSON path syntax, execution branching, expression syntax) assumes technical comfort. Non-technical marketing or ops teams take 2-4 weeks to get productive vs Zapier's 2-4 days. If your automations are owned by a non-technical team, Zapier or Make's gentler curves can be worth the higher cost.
Cloud SLA and enterprise support still maturing. Compared to Zapier and Workato, the cloud tier's enterprise support story is younger. Business plan (€800/mo) includes SSO, RBAC, log streaming, and 99.9% SLA, but custom DPAs, multi-year contracts, and procurement defaults are less battle-tested than Zapier's enterprise sales motion. Improving fast (Series B funded the build-out) but if you need full enterprise procurement machinery on day one, this is a friction point.

ICP fit — verdict by use case

Verdict	Use case	Why
✅ Yes	Engineering-led ops or RevOps team running automation as code	Code-extensible nodes + execution model fit how engineering teams already think about pipelines.
✅ Yes	Team hitting Zapier per-task pricing walls (>$300/mo)	The execution-vs-task math compounds. Migration usually pays back in the first quarter.
✅ Yes	Compliance-heavy industries needing self-host or data sovereignty	Self-host is a structural advantage no per-task SaaS can match.
✅ Yes	Team building AI agent workflows (research, content, support triage)	Native AI agent nodes + vector store integrations + chain composition = competitive with LangChain.
✅ Yes	Operator with mid-volume polling-heavy automations on self-host	No execution cap on self-host means polling triggers are free. Webhooks where possible, polling where needed.
⚠️ Maybe	Team running 5-15 simple workflows with no growth ahead	Zapier or Make is fine if you'll never scale. n8n's advantages mostly compound at volume.
⚠️ Maybe	Hybrid team: technical lead, non-technical operators	Works if the technical lead owns the harder workflows and operators build simple ones. Otherwise the curve hurts.
❌ No	Non-technical marketing or ops team with no engineering support	Zapier or Make are structurally better fit. n8n's curve costs more in time than n8n saves in money.
❌ No	Need integration with a niche tool not in n8n's catalog or HTTP-accessible	If the long-tail integration is mission-critical, Zapier's 6,000-integration catalog wins.
❌ No	Enterprise procurement requires day-one SLAs, custom DPAs, multi-year contracts	Workato and Zapier Enterprise have the procurement machinery. n8n is closing the gap fast but isn't there yet.

Switching cost — what migration actually takes

Step	Time
Cloud signup + plan selection	< 30 min
Self-host setup (Docker + reverse proxy + HTTPS)	1-3 hrs
Database migration (SQLite → Postgres for production)	1-2 hrs
Workflow rebuild (per workflow, manual reconstruction)	~ 2-4 hrs
Workflow rebuild for complex multi-step (custom logic)	~ 4-8 hrs
Authentication migration (OAuth, API keys per integration)	~ 30 min/integration
Test + production cutover (parallel run for 1-2 weeks)	Real risk window
Self-host upgrade cadence (quarterly minor, plus security)	~ 30 min/quarter ongoing

Real total: a 30-workflow Zapier-to-n8n migration runs 60-120 hours of focused work spread over 2-6 weeks, including the parallel-run period. Cost savings usually pay back the migration time within the first quarter — often the first month at meaningful volume.

How n8n compares to the field

Tool	Headline	Honest take
Zapier	Most popular, biggest catalog, per-task pricing	6,000+ integrations and the easiest UX in the category. The trade-off is per-task pricing that scales linearly with workflow complexity. For teams running 5-15 simple workflows under 5,000 tasks/mo, Zapier is fine. Above that, n8n's pricing structure wins by 5-10x at equivalent volume. Use Zapier if you need niche integrations or a non-technical team owns automations; use n8n if you're scaling.
Make (Integromat)	Visual middle ground, per-operation pricing	More polished UX than n8n, more powerful than Zapier, ~1,500 integrations. Per-operation pricing is similar trap to Zapier's per-task. Reasonable choice for SMBs that prefer SaaS polish and don't plan to scale to high-volume automation. n8n wins for technical teams, scale, or self-host requirements; Make wins on UX polish at SMB scale.
Workato	Enterprise iPaaS, $10-50K/yr+ contracts	Different category. Workato is enterprise iPaaS with deep integration depth, governance features, and procurement machinery — used by Fortune 500 IT teams. n8n is a fraction of the cost but doesn't have the same enterprise sales motion or compliance certifications (yet). For mid-market and below, n8n wins on cost-to-capability ratio. For Fortune 500 IT-led integration projects, Workato is still the default.
Zapier Tables / Interfaces	Zapier's expansion into databases + UI	Tangentially related — Zapier is bundling a low-code database and UI layer to expand beyond pure automation. n8n doesn't compete here directly. If your use case is "Zap + a UI to capture form data + a small DB", Zapier's bundle is convenient. If your use case is workflow execution at scale, the DB/UI bundle is irrelevant and n8n wins on the automation side.
Tray.io	Mid-market iPaaS, embedded in product motion	Tray sits between Workato and Zapier — focused on embedded automation for SaaS companies (offering Tray-powered integrations to your customers). Different go-to-market than n8n. If you're building customer-facing integration capabilities into a SaaS product, Tray's embedded model fits. For internal automation, n8n is significantly cheaper.
Pipedream	Code-first automation, generous free tier	Closer competitor to n8n than the no-code tools. Code-first orientation, strong API integration, generous free tier (10k credits/mo). Pipedream is more code-centric and less visual; n8n splits the difference better with a visual editor that supports inline code. For teams that want pure code, Pipedream is reasonable; for teams that want visual + code, n8n wins.

For the head-to-head decision most buyers actually face, see our n8n vs Zapier vs Make three-way comparison.

Self-host vs cloud — how to decide

The first decision every n8n buyer faces. Here's the cleanest decision tree:

Choose self-host if: you have any DevOps capacity (or someone comfortable with Docker basics), you're running polling-heavy workflows, you have data sovereignty / compliance requirements, or you're running >10k executions/mo and want to control infrastructure cost.
Choose cloud if: you have zero DevOps capacity, you're running <10k executions/mo (mostly webhook-driven), you want vendor support and SLAs without negotiating, or you're proving the tool before committing infrastructure.
Hybrid pattern (common): start on cloud to validate, then migrate to self-host once you cross the volume threshold where cloud pricing exceeds infra cost (~10-15k executions/mo for most teams). The cloud → self-host migration is straightforward — workflows export and import directly.

What G2 + Capterra reviewers consistently say

Reviewers consistently call out the per-execution pricing as "the reason we switched from Zapier."
Self-host comes up repeatedly as a compliance unlock — financial services and healthcare in particular.
AI agent capabilities are the most-cited recent addition; reviewers compare favorably to LangChain for non-developers.
Documentation gaps and vague error messages are the most-cited frustrations across G2 and Capterra.
Setup time is the second-most-cited frustration — typically 1-3 hours for self-host, 1 day for non-technical users on cloud.

Aggregate themes from public reviews on G2 and Capterra. Individual reviews vary; this is what we see consistently across hundreds of operator opinions in the category.

Common use cases we see

Lead enrichment + CRM sync pipelines: fetch from Apollo / Clay / Clearbit, enrich with LLM-driven inference, dedupe, sync to HubSpot or Salesforce. Replaces a $200-500/mo Zapier setup with a $10-30/mo n8n one.
AI content / research pipelines: RSS or webhook trigger → LLM-driven research with vector store retrieval → draft content → human review queue. The 2026 use case n8n is winning fastest.
Internal tool automation: CI/CD webhook glue, Linear → Slack notifications, GitHub PR triage, deployment status broadcasting. Engineering teams build dozens of these and hate paying Zapier per task for them.
Customer data orchestration: sync between Stripe / billing → CRM → product database → marketing automation, with custom transformation logic that would otherwise live in a backend microservice.
Compliance-sensitive data flows: healthcare, financial services, government — anywhere data sovereignty matters and SaaS automation tools fail the procurement review. Self-host is the unlock.

FAQ

Is n8n actually free?: The Community Edition (self-hosted) is fair-code licensed and free with unlimited executions. You provide the infrastructure — typically $5-20/mo for a small DigitalOcean, Hetzner, or AWS Lightsail instance, plus your own time to maintain it. The cloud-hosted plans start at €24/mo (Starter, 2,500 executions) and scale to €60/mo (Pro, 10,000) and €800/mo (Business, 40,000). Self-host wins on cost above ~10k executions/mo if you have any DevOps capacity; cloud wins on convenience below that.
How does n8n actually compare to Zapier on cost?: The structural difference is execution counting. n8n charges per workflow execution (one full workflow run). Zapier charges per task (every step is a task). A 10-step workflow that runs 1,000 times = 1,000 n8n executions but 10,000 Zapier tasks. At Zapier's $0.025-0.05/task pricing, that's $250-500/mo on Zapier vs ~$24-60/mo on n8n cloud or ~$10/mo on n8n self-host. The cost gap compounds with workflow complexity, which is why teams running multi-step automations migrate quickly once they do the math.
Do I really need a developer to run self-hosted n8n?: You need someone comfortable with Docker basics or basic cloud server management — that's a 1-3 hour setup, not a multi-week project. Once running, the visual workflow editor is accessible to non-developers. The two real DevOps tasks are: (1) initial install on a VPS with HTTPS and a reverse proxy, (2) ongoing updates and backups (~30 min/quarter). If your team has zero DevOps capacity, n8n cloud (€24/mo Starter) skips this entirely and you still get most of the cost advantage vs Zapier.
What about polling triggers eating my execution budget?: This is the most common pricing surprise on n8n cloud. A polling trigger checking every 5 minutes runs 12 times/hour × 24 × 30 = 8,640 executions/month — that single workflow exhausts the Starter plan (2,500/mo) by day nine and burns most of the Pro plan (10,000/mo). Solutions: (1) use webhooks instead of polling wherever possible, (2) reduce poll frequency to hourly or daily for non-time-critical flows, (3) run polling-heavy workflows on self-hosted (no execution cap), (4) consolidate multiple polls into a single scheduled workflow that branches.
How does n8n compare to Make (formerly Integromat)?: Both are visual workflow automation tools. Make is the middle ground — more polished UX than n8n, more powerful than Zapier, ~1,500 integrations. n8n has fewer native integrations (~1,000) but its HTTP node + custom code support means it can talk to anything with an API. Make charges per operation (similar trap to Zapier's per-task model); n8n charges per execution. For SMBs that prefer a polished SaaS UX, Make is reasonable. For technical teams or anyone planning workflows that scale, n8n's pricing model is structurally better.
What integrations does n8n actually support?: About 1,000 native integration nodes covering the major SaaS tools (Slack, Gmail, Salesforce, HubSpot, GitHub, Notion, Stripe, Shopify, Postgres, MySQL, MongoDB, etc.) plus generic HTTP request, webhook, and code execution nodes that connect to anything with an API. The catalog is smaller than Zapier (~6,000) and Make (~1,500), but the HTTP + code nodes close most of the gap for teams comfortable reading API docs. The integration depth varies — some integrations are surface-level OAuth wrappers, others have dozens of well-modeled actions.
Is n8n suitable for compliance-heavy industries?: Yes — self-hosting is one of the main reasons regulated industries choose n8n. Your data stays on infrastructure you control, which simplifies SOC2, HIPAA, GDPR data residency, and ITAR-style requirements. The fair-code license permits self-host for internal use without restriction. n8n cloud is also SOC2 Type 2 certified for the cloud-hosted tier. If your buyer requires data sovereignty, n8n is structurally better positioned than any per-task SaaS tool that must process your data on their infrastructure.
How well does n8n handle AI agent workflows in 2026?: This is where n8n has become a serious player in the last 12 months. The AI agent layer ships first-class support for multi-step agent workflows, native nodes for OpenAI / Anthropic / Google Gemini, vector store integrations (Pinecone, Qdrant, Weaviate), and chain composition that's closer to LangChain than to Zapier's natural-language workflow builder. For teams building agent-style automations (research workflows, content pipelines, support triage), n8n is competitive with purpose-built agent frameworks while staying inside a visual workflow editor.
Can n8n replace Workato or Zapier outright?: For most use cases, yes. The migration overhead depends on workflow complexity and integration coverage. Teams running 30-50 active Zapier automations typically take 2-6 weeks for a clean migration if the integrations are all in n8n's catalog. If you depend on long-tail integrations (the 5,000+ niche apps Zapier supports that n8n doesn't), the migration includes building those connections via HTTP nodes. Cost savings usually pay back within the first quarter — often the first month at scale.
What's the catch — what does n8n not do well?: Three real weaknesses. (1) Documentation can be thin or out of date for newer features; community forum and Discord fill the gap but expect to dig. (2) Error messages are sometimes vague ("Problem executing workflow") and require manually inspecting the canvas to find the failed node — frustrating during initial debugging. (3) The learning curve is real for non-technical users — concepts like webhook authentication, JSON traversal, and execution branching assume some technical comfort. If your automations are owned by a non-technical marketing or ops team without engineering support, Zapier or Make's gentler curves win.