n8n
Visual and open-source workflow orchestration platform allowing you to connect APIs, SaaS systems and databases without code or with punctual code nodes. Alternative to Zapier and Make with two critical enterprise advantages: self-host option (data never leaves your infrastructure) and per-execution pricing model much more economical at high volume. It's the de facto standard in enterprise mid-market automation in 2026.
What n8n is and why it's displacing Zapier/Make in 2026
n8n is a workflow orchestration platform allowing you to connect systems (APIs, SaaS, databases, own systems) via visual nodes and punctual code when needed. Competes in the same space as Zapier and Make (formerly Integromat) but has two decisive advantages for enterprise: open-source with self-host option, and its per-execution pricing model is 3-10x cheaper at high volume.
Mid-market Spanish adoption exploded in 2025-2026 for three concurrent reasons: (1) Zapier raised prices aggressively between 2023 and 2025, (2) regulated companies need on-premise data Zapier/Make don't offer, and (3) AI teams adopted it to orchestrate agents and LLM flows — n8n has native nodes for OpenAI, Anthropic, LangChain, embeddings and vector DBs.
For serious enterprise processes (ERP integrations, workflows with regulated data, AI agent orchestration), n8n is today the default choice unless specific reasons for choosing alternative.
n8n isn't a replacement for all automation types. For massive ETL data flows use Airflow or Prefect. For full custom development, go direct to code. n8n is the sweet spot between "no code" and "full code" — most enterprise processes fit here.
n8n vs Zapier vs Make vs Airflow: when each
Zapier remains the best option for very small SMEs or quick pilots that won't scale in volume. UX is simpler and the community huge.
Make is competitive with n8n in visual workflow power but doesn't offer self-host, ruling out many regulated enterprise cases.
Airflow is the standard in data engineering for complex pipelines, but demands high technical profiles (Python, DAGs, DevOps). Not a realistic alternative for orchestrating business processes operated by non-technical teams.
| Criterion | n8n | Zapier | Make | Airflow |
|---|---|---|---|---|
| Price at high volume | ★★★★★ | ★★ | ★★★ | ★★★★★ |
| Self-host / on-prem | Yes (native) | No | No | Yes |
| Visual interface | Yes (powerful) | Yes (simple) | Yes (powerful) | No |
| Native AI nodes | ★★★★★ | ★★★ | ★★★ | ★★ |
| Workflow complexity | High | Low | High | Very high |
| Learning curve | Medium | Very low | Medium | High (Python) |
| Community | ★★★★ | ★★★★★ | ★★★★ | ★★★★★ |
| Fit for mid-market enterprise | ★★★★★ | ★★★ | ★★★★ | ★★★ |
Architecture and deployment options
n8n offers three deployment options with different price, security and operational implications.
- n8n Cloud — SaaS operated by n8n. Zero DevOps, high availability, automatic updates. Data flows through n8n's infrastructure in EU/US per chosen region. Very competitive pay-per-execution model up to ~100,000 executions/month.
- Self-hosted community edition — Install in your infrastructure (Docker, k8s, VPS). Free, no execution limit, you operate. Standard for companies with regulated data or very high volume. Requires DevOps profile to operate well.
- Self-hosted enterprise edition — Commercial n8n license with SSO, granular RBAC, complete audit logs, horizontal scaling, formal support and SLAs. For large companies needing formal governance. Price on demand (€5-30k/year range).
Self-host community edition seems "free" but has real operating cost: DevOps (setup, updates, monitoring), infrastructure (compute + Postgres DB + Redis), backups and recovery. Budget €300-800/month per volume for serious self-host in production.
Real use cases in Spanish enterprise
n8n is horizontal — applied to any sector with systems to connect. These are cases where we see most impact in Spanish enterprise 2026.
- AI agent orchestration — n8n triggers agents, manages retries, timeouts and human escalation. It's layer 4 of the agent architecture: process brain while the LLM thinks.
- CRM ↔ ERP ↔ marketing integration — new lead enters via form → AI qualification → CRM creation → send to marketing automation → sales follow-up. All without manual exports.
- Back-office automation — invoice arrives by email → OCR + validation → ERP creation → approver notification → accounting entry. Reduces invoicing time 60-80%.
- Sync between scattered SaaS systems — change in Salesforce automatically reflects in HubSpot, Slack, internal Google Sheets and proprietary system. Eliminates duplicated "single truth".
- Lightweight data pipelines — daily extraction from multiple sources (APIs, DBs, files) → transformation → BI load. Alternative to Airflow for medium-complexity pipelines.
- Chatbot with real backend — the conversational bot (channel) calls n8n to execute actions (change address, cancel order, check stock) with controlled real system access.
Real cost: cloud vs self-host
Rule of thumb: self-host becomes profitable from ~50,000 executions/month, provided you have DevOps profile to operate it. Below, Cloud is better. Above 500k executions/month, self-host is clearly cheaper and profitable even paying enterprise for SLA and support.
| Monthly volume | n8n Cloud | Self-host community | Self-host enterprise |
|---|---|---|---|
| < 5,000 executions | €20-50/mo | €300-500/mo (DevOps) | doesn't pay off |
| 5,000-50,000 | €50-250/mo | €400-800/mo | doesn't pay off |
| 50,000-500,000 | €250-1,500/mo | €600-1,500/mo | €600-2,000/mo |
| 500,000-5M | €1,500-8,000/mo | €1,000-3,000/mo | €1,500-4,000/mo |
| > 5M | not competitive | €2,000-6,000/mo | €3,000-8,000/mo |
Security, compliance and best practices
- Encrypted credentials at rest — n8n uses AES-256. Configure encryption keys outside the container (env vars, K8s secrets). Don't hardcode.
- Periodic credential rotation — automate with Vault, AWS Secrets Manager or similar. Manual = risk.
- Granular RBAC — enterprise edition offers roles and permissions per workflow. In community, use separate organisations per domain (marketing, ops, finance) with distinct instances if risk justifies.
- Audit logs — who edited what workflow and when. Enterprise does it natively. Community requires external integration (Loki, Datadog).
- Regulated data never in Cloud — for banking (DORA), health (GDPR art. 9), insurance (DGSFP), only viable option is self-host on-premise or in European cloud with explicit contract.
- Automatic backup of workflows and credentials — daily export + git versioning. A broken workflow without backup = hours of rebuild.
- Separate environments — dev, staging, production as distinct instances. Never edit directly in production.
How to start in 4 weeks
- Week 1 — Select a single concrete process (ideal candidates: CRM-marketing sync, internal notifications, form ingest). Install n8n Cloud (trial) or Docker locally to prototype.
- Week 2 — Design the complete workflow with visual nodes. Don't add error handling or edge cases yet — validate that the main logic works with real data.
- Week 3 — Add error layer: try/catch on critical nodes, retries with backoff, Slack notification on failure, manual fallback. Dependencies documentation.
- Week 4 — Deploy to production with initial supervision. Monitor for 2 weeks before leaving on auto. Document maintenance procedure.
Always start with Cloud if the process doesn't handle regulated data. Zero DevOps, 100% focus on logic. When volume or compliance justifies, migrate to self-host — the migration is direct (workflow export/import).
Frequently asked questions
Is it worth switching to n8n if we already use Zapier?
Depends on volume and sector. If you're in a very small SME with less than 3,000 tasks/month and don't handle regulated data, Zapier is still simpler and you won't gain much migrating. Above 5,000 tasks/month, n8n Cloud saves you between 30% and 70% of cost with fewer "steps per workflow" limitations.
If you're in mid-market or enterprise with sensitive data, n8n is practically mandatory: Zapier doesn't offer self-host and its enterprise contracts are expensive with less flexibility.
Typical migration: 2-6 weeks depending on number of workflows to migrate. Operating cost recovers in 3-9 months.
Does n8n comply with GDPR and AI Act if we automate processes with personal data?
Yes, as long as you deploy correctly. With self-host in your infrastructure or in European cloud with Spain residency, personal data never leaves your perimeter and GDPR complies with the same guarantees as any other internal system.
For high-risk processes under AI Act (HR, credit, biometrics), the n8n workflow is part of the registered AI system and you must document: what data it processes, what AI models it invokes, what automated decisions it makes, what human oversight exists.
As with any system, you must do impact analysis (DPIA) when processing requires it and sign processing agreements with providers (Anthropic, OpenAI, etc.) if the workflow sends personal data to LLM APIs.
Can I orchestrate AI agents with n8n or is something more specific needed (LangChain, LangGraph)?
n8n orchestrates AI agents perfectly for most enterprise cases. Has native nodes for Anthropic, OpenAI, LangChain, embeddings and vector DBs. Its strength is the orchestration part (retries, timeouts, escalation, human in loop) — not the agent "reasoning", which still lives in the LLM.
LangChain and LangGraph are better when the agent itself is very complex (multiple sub-agents, complex memory, state graphs with many branches). Many times the optimal architecture is hybrid: n8n orchestrates the global flow and calls an agent built with LangChain/LangGraph on heavy reasoning steps.
n8n's advantage: operations teams can see and modify the workflow. With pure LangChain, only a developer can touch it.
What if the n8n vendor disappears or changes business model?
n8n is open-source (Sustainable Use License) with self-host available at no cost indefinitely. Even if the company disappeared, you'd keep being able to operate your self-hosted instance. Real vendor lock-in is low.
Additionally, workflows export to JSON and are portable. Migrating to another platform is painful but feasible — you're not trapped.
The company behind n8n (founded in Berlin in 2019) has stable funding, very active community and growing enterprise presence. Discontinuity risk is low compared to smaller providers in the space.