How n8n Became the Fair-Code Automation Layer for the Agentic Era

Why n8n matters right now

n8n is the workflow automation platform for technical teams that want visual building without giving up code, and want AI agents without giving up control. In May 2026, SAP made a strategic investment in n8n that more than doubled its valuation to $5.2 billion from $2.5 billion a year earlier, with n8n now embedded inside Joule Studio on SAP Business Technology Platform (Pulse 2.0, CXO Digital Pulse, Verdict). That deal tells the same story n8n’s customers already tell us: when AI agents have to touch real systems, the plumbing has to be visible, debuggable, and yours to host.

Founder and CEO Jan Oberhauser put it plainly in the announcement:

“For n8n, securing SAP as a strategic investor marks a pivotal moment. As one of the world’s largest enterprise software companies, its decision to back n8n and to embed us inside Joule Studio reflects genuine confidence in our platform and our vision.”

We have been tracking n8n since the Sequoia seed round in 2020 and the Series A in 2021. The 2026 product is a different beast than the original node-painting tool. It now ships native AI agent nodes, a first-class Model Context Protocol (MCP) Client and MCP Server Trigger, Evaluations for production AI, and 1,500+ pre-built integrations while still being source-available on GitHub. The headline metric: 199k stars and 59.8k forks as of this writing, with a 4.7/5 average on G2 and roughly 1.7 million monthly active developers and builders according to the company’s SAP press materials.

We will dig into the numbers, the AI story, the security posture, the pricing, the founder’s positioning, and where we think n8n actually fits. If you are evaluating n8n in 2026, this is the briefing we wish we had.

What is n8n, really?

n8n is a fair-code workflow automation platform that combines a visual node-based canvas with custom JavaScript and Python, 1,500+ pre-built integrations, and a self-hostable Community edition alongside managed cloud plans (GitHub, n8n.io). The “fair-code” part matters: the code is visible on GitHub under the Sustainable Use License, which lets you run, fork, and modify n8n for your own automations, but does not let you resell it as a hosted service. Think of it as “open-core with a clear line around commercial competition.”

In a 2019 README, Oberhauser explained the name (GitHub): “node-” for the node-based view and the Node.js stack, “-mation” for automation, compressed to n-eight-n. Pronounce it the same way you would read the spelling.

Today, n8n describes itself as “the platform for AI agents and workflow automation,” with the explicit positioning that it is an automation engine first and an AI surface second (n8n.io). That framing shows up in how the product treats AI: the AI Agent node is a regular node on the canvas, surrounded by deterministic logic, human-in-the-loop checkpoints, and branch nodes that route on real data.

n8n at a glance (2026)

Metric	Value	Source
Valuation	$5.2B (May 2026, post-SAP)	Pulse 2.0, CXO Digital Pulse
Prior round valuation	$2.5B (Series C, late 2025)	Tracxn
Series B	€55M led by Highland Europe (Mar 2025)	EU-Startups
Series A	$12M (Apr 2021)	TechCrunch
Seed	Sequoia (Mar 2020)	TechCrunch
Enterprise customers	1,400+	Pulse 2.0
Monthly active developers/builders	~1.7 million	Pulse 2.0
Integrations	1,500+ (1,000+ pre-built + HTTP/GraphQL/webhook)	GitHub, n8n.io
GitHub stars	199k (59.8k forks)	GitHub
Workflow templates	10,000+	n8n.io
Cloud hosting region	EU (Microsoft Azure, Frankfurt)	n8n Security
Compliance	SOC 2 audited, SOC 3 public, GDPR-aligned	n8n Security
Founded	2019 in Berlin, Germany	n8n Press

The story so far: 2019 to 2026

The company was founded in Berlin in 2019 and quickly attracted seed funding from Sequoia (TechCrunch, 2020). A $12M Series A followed in April 2021 (TechCrunch). In March 2025, n8n closed a €55M Series B led by Highland Europe, with participation from HV Capital and existing investors Sequoia, Felicis, and Harpoon (EU-Startups). At the time, the company said it had crossed 200,000 active users and grown ARR 5x in the previous year.

Then the AI wave hit. By late 2025, a Series C had pushed n8n’s valuation to $2.5 billion (Tracxn). GetLatka’s 2025 estimate placed ARR at roughly $40M against the $2.5B mark, with ARR growth reported at multiples of the prior year (GetLatka). In May 2026, SAP’s strategic investment doubled the valuation to $5.2 billion and embedded the platform inside Joule Studio on SAP BTP (Pulse 2.0, Tech.eu). SAP’s CEO Christian Klein framed the deal as making agentic AI work in the real world: “To provide accurate and secure business outcomes at scale, agentic AI must be grounded in deep process knowledge, reliable data, and enterprise-grade governance.”

A second signal of the platform’s gravity: n8n is now used by 34% of Fortune 500 enterprise teams for advanced security and DevOps workflows, per the n8n Enterprise page. Logos on that page include Microsoft, Meta, NVIDIA, Deutsche Telekom, Novo Nordisk, Vodafone, Dell, Cummins, Mercedes-Benz, and Mistral AI.

What n8n actually does in 2026

The core product is a node-based workflow editor. You drag nodes onto a canvas, wire them together, and the platform runs the graph when a trigger fires. Where n8n distinguishes itself is the same set of choices it has made since 2019, plus a lot of new AI surface area.

1. Visual canvas with code at every step

Every node exposes its inputs and outputs on the canvas. You can map fields with a UI or write JavaScript or Python expressions inline. The Code node lets you drop full snippets of either language into the workflow. The HTTP Request, GraphQL, and Webhook nodes handle anything that does not have a first-party connector.

2. Native AI agent nodes

n8n ships LangChain-powered AI nodes for chat models, embeddings, memory, vector stores, and tools. The AI Agent node is the centerpiece: you wire it to any chat model, attach memory and tools, and it executes inside the same canvas as your deterministic logic. You can choose any provider (OpenAI, Anthropic, Google, open-source models via Ollama) and switch providers without rebuilding the graph.

3. First-class MCP support

MCP (Model Context Protocol) is the open standard for letting AI agents call external tools. n8n has both a MCP Client Tool node and an MCP Server Trigger. The Client lets your workflow call any MCP-enabled tool; the Trigger turns any n8n workflow into an MCP server that agents like Claude or Lovable can call. That second one is the interesting move: it makes n8n a programmable surface for other agents.

4. Evaluations for production AI

The platform includes an Evaluations feature that scores AI workflows against a defined test set. You can compare models, prompts, and tool configurations with hard numbers, then monitor drift after deployment. For teams that have been burned by agents that look great in demos and fail quietly in production, this is the part of the roadmap that matters most.

5. Self-hostable from day one

The Community Edition on GitHub is free and includes the full editor. You can npx n8n to try it locally, run it via Docker, or deploy on a VPS. Paid self-hosted tiers (Business and Enterprise) add collaboration, multiple environments, Git-based version control, and SSO. For many teams, this is the entire reason to choose n8n over a closed alternative.

6. AI Workflow Builder and Chat Hub (cloud)

The AI Workflow Builder describes a workflow in plain English and ships a working draft, with iterative chat-based refinement. The Chat Hub acts as a company-wide control plane: every employee gets a chat surface that can route to any configured agent or workflow, with permissions inherited from the underlying workflow.

How n8n stacks up: a 2026 comparison

Automation buyers in 2026 are usually choosing between closed SaaS, fair-code, and pure open source. The table below compares the four most common benchmarks for n8n based on the company’s own vs. Zapier and vs. Make pages, plus public product documentation.

Dimension	n8n	Zapier	Make	Workato / Tray
Pricing model	Per full execution, unlimited steps	Per task (every step counts)	Per operation + data transfer	Per recipe/connector (enterprise)
Unlimited users	Yes on all plans	Enterprise only (Team capped at 25)	Yes on Teams+	Yes (enterprise)
Self-hosting	Free Community Edition + paid Business/Enterprise	No	No (on-prem agent for data only on Enterprise)	Workato on-prem; Tray cloud only
Custom code	JS and Python at any step	Code by Zapier (JS/Python, sandboxed, 30s/256MB limits)	Custom JS on Enterprise only	Yes (enterprise)
Open source	Source-available (Sustainable Use License)	No	No	No
AI features	AI Agent node, MCP client/server, Evaluations, AI Assistant	AI actions, AI Copilot (beta)	Limited; expanding	Enterprise AI add-ons
Public integrations	1,000+ pre-built (1,500+ with HTTP/GraphQL/community)	8,000+	2,800+	1,000+ (enterprise-curated)
Native branching	IF, Switch, Merge, Loop Over Items	Paths (premium)	Router (premium)	Yes (enterprise)
Best for	Technical teams, AI agents, regulated workloads	Non-technical teams, simple Zaps	Mid-market, visual scenarios	Large enterprises, IT-led integration

Three takeaways from that table:

1. Pricing discipline is n8n’s structural advantage. Charging per execution instead of per step means a 200-step AI workflow costs the same as a 2-step webhook. n8n’s own execution-advantage explainer estimates that workflows equivalent to roughly 100,000 tasks can cost over $500 per month on competitors, versus about $50 on n8n’s Pro plan. That claim is on n8n’s blog, so treat it as marketing, but the underlying arithmetic is correct.
2. Closed platforms still win on connector breadth. Zapier’s 8,000+ app library is unmatched for non-technical users. n8n closes the gap with community-built nodes and universal HTTP/GraphQL/webhook connectors, but if you need a niche SaaS connector and cannot write 20 lines of HTTP, the closed tools have an edge.
3. Open source is the moat, but it is not a feature. Activepieces, Windmill, Temporal, and Kestra are all credible open alternatives. n8n wins on AI agent orchestration and on the depth of its enterprise governance story, not on the license alone.

n8n vs Zapier at a glance

The pitch n8n itself makes (n8n.io/vs/zapier): Zapier is the right tool for simple, linear automations owned by non-technical users. n8n is the right tool when complexity, code, data residency, or AI agents are in the mix. We agree, with the caveat that Zapier’s “AI Copilot” and a long list of pre-built AI actions have closed part of the gap for shallow agent use cases.

Pricing in 2026: what you actually pay

n8n’s published Cloud pricing (EUR-listed, billed annually) is structured around executions and AI credits:

• Starter: €20/month, 2,500 executions, 5 concurrent executions, 2,300 AI credits, 1 shared project, forum support. Aimed at individuals and small experiments.
• Pro: €50/month, 10,000 executions, 20 concurrent executions, up to 13,700 AI credits, 3 shared projects, admin roles, global variables, workflow history, and execution search.
• Business: €667/month, 40,000 executions, 6 shared projects, SSO/SAML/LDAP, multiple environments, Git version control, scaling options. Self-hosted only.
• Enterprise: Custom pricing, 200+ concurrent executions, unlimited projects, 365 days of insights, external secret store integration, log streaming, dedicated support with SLA, and invoice billing. Available as cloud or self-hosted.

A few pricing mechanics that we have read carefully because they tend to surprise buyers:

• Executions are the unit, not steps. A 2-step workflow and a 200-step AI agent both count as one execution when they run from start to finish (n8n pricing FAQ).
• AI Assistant is in early preview on cloud. Credits refresh monthly and do not roll over. You cannot buy more credits yet; if you outgrow your tier, the answer is to upgrade.
• Business overages. 300,000 extra executions on Business are billed at €4,000 per bucket. The sales team reaches out when you cross 80% of your annual quota.
• Free trial. Starter and Pro trials need no credit card. Business trial does.
• Self-hosting costs you infrastructure, not the software. The Community Edition is free; Business and Enterprise self-hosted add collaboration, SSO, and governance features that you would otherwise buy from a closed vendor.

For self-hosted buyers, the total-cost-of-ownership question is honest: you save license cost but you take on container orchestration, database management, upgrades, and incident response. We will return to this in the limitations section.

The founder’s position in his own words

Jan Oberhauser is not a typical SaaS CEO. He codes. He maintains the original “What does n8n mean?” README block. He shows up in the community forum. In the 2025 Series B announcement, he framed the company’s mission this way (EU-Startups):

“Automation shouldn’t be a black box – companies need transparency, customisation and cost efficiency. With n8n, we’ve built more than a platform; we’ve built a community that loves and relies on us. From individual contributors to global enterprises, n8n gives everyone the powers of a 10x developer which is vital given the explosion of AI in the workplace.”

Three things to notice in that statement. First, the explicit refusal of “black box” — n8n is not trying to be the easiest tool, it is trying to be the most inspectable. Second, the word “community,” which is not a marketing flourish at n8n: the community forum has tens of thousands of active members, and community events on Luma draw hundreds of builders to local meetups from Ibadan to Tallahassee. Third, the explicit connection between automation and AI agents, which is the strategic thesis behind the SAP deal and the new product surfaces.

The 2026 AI feature set, in plain English

If you have not looked at n8n since 2024, the AI surface area is unrecognizable. Here is what the product does today, and where each feature is documented.

AI Agent node and LangChain integration

The AI Agent node wires an LLM into a workflow, attaches tools, and lets the model decide which steps to call. The same node also supports multi-agent setups where a “tools agent” dispatches to specialized sub-agents for research, writing, QA, and data analysis. Any LLM provider works: OpenAI, Anthropic, Google, Mistral, and local models through Ollama.

MCP Client and MCP Server Trigger

MCP is the open standard for tool calling between agents. n8n has both ends of the wire:

• The MCP Client Tool lets a workflow call any MCP-enabled tool, including tools exposed by other n8n instances.
• The MCP Server Trigger exposes a workflow as an MCP server, so external agents (Claude, Lovable, custom agents) can call your automation as a tool.

This is the part of the roadmap that turns n8n from “automation tool” into “agent infrastructure.” It also explains why SAP embedding n8n inside Joule Studio matters: SAP customers can now compose agents that orchestrate across SAP systems plus everything else an n8n workflow can reach.

Evaluations

Evaluations are how you stop guessing whether your AI workflow is good. You define a test set, the platform runs your workflow against each case, scores the output, and reports pass/fail by metric. This is the part of the n8n roadmap that maps directly to the OWASP top 10 LLM risks — specifically excessive agency, unbounded consumption, and improper output handling.

Guardrails, human-in-the-loop, and AI governance

n8n’s AI governance page and LLM security guide both emphasize that production AI needs to be anchored in deterministic steps and human approvals. The platform ships guardrails nodes, manual approval nodes, and rate-limit patterns that keep agents from going off the rails.

AI Workflow Builder and Chat Hub

AI Workflow Builder describes a workflow in plain English and returns a working draft, which you then refine through chat. The Chat Hub gives every employee a chat surface that can route to any configured model or agent workflow, with permissions inherited from the underlying workflow.

Who is actually shipping with n8n in 2026?

The case studies on n8n.io/case-studies are unusually concrete. Three that capture different buyer profiles:

Huel: AI-first company culture, nine months in

Huel, the global nutrition company, moved to n8n Enterprise and built 200 live workflows in nine months, with 100+ employees actively using n8n. The team saved over 1,000 hours of manual work and £100,000 in annual software costs by replacing several single-purpose SaaS tools with custom n8n workflows. CTO Ollie Scheers is direct about why: “n8n was the big unlock. Tools like ChatGPT and Claude are great, but n8n is the thing that allows you to integrate AI into your work and your processes in a safe and controlled way.”

Vodafone: SOAR plus workflow automation in one tool

Vodafone UK chose n8n over IBM Resilient and Tines to handle both SOAR (security orchestration, automation, and response) and broader workflow automation. Since August 2024, the team has launched 33 workflows that process 3 to 5 billion security events per month, save 5,000 person-days of manual work, and avoid £2.2 million in costs, with continued savings of roughly £300k per month in 2025 (Vodafone case study). The team has since started exploring agentic SOC patterns for the next phase.

Musixmatch: data retrieval for a global lyrics platform

Musixmatch built 27 custom modules on n8n to handle client data requests without involving engineers. Time to complete a client request dropped to 15 seconds, and the team saved 47 days of engineering work in the first four months (Musixmatch case study).

Two more that round out the buyer profile:

• Fullscript, a healthcare SaaS company, scaled n8n to 200+ employees (over 30% of the org), produced 23 production workflows, and built a security investigation workflow in two days that saves 3,600 engineering hours a year and cut security investigations by 97%.
• SanctifAI, a Human-AI collaboration platform, spun up its first n8n workflow in two hours — 3x faster than writing Python controls for LangChain — and now orchestrates tasks across 400+ workforces.

Security, compliance, and on-prem posture

The n8n security page is unusually detailed for a vendor of this size. The headline posture:

• SOC 2 audited, with a public SOC 3 report downloadable. SOC 2 reports are available to enterprise customers under NDA.
• GDPR-aligned, with EU data residency for cloud customers via Microsoft Azure in Frankfurt, Germany.
• Encryption at rest using Azure Storage server-side encryption (AES-256, FIPS 140-2 compliant).
• OAuth-first for third-party app connections, with credentials encrypted in a per-instance secret store and never logged by default.
• RBAC and SSO (SAML, LDAP) on paid plans; MFA enforced where supported.
• Audit logs retained for at least 12 months, with the last three months immediately queryable. Enterprise plans stream to your SIEM.
• Third-party vulnerability scans at least every 90 days and penetration tests at least annually.
• Bug-bounty and vulnerability disclosure program.

Two caveats we want to flag, both for transparency:

• Self-hosted customers own encryption at rest. n8n salts and hashes passwords, but disk encryption, key management, and TLS termination are your responsibility. The security docs spell this out plainly.
• n8n had a real security incident in early 2026. CISA added a critical n8n vulnerability to its Known Exploited Vulnerabilities catalog, and an earlier unauthenticated server takeover flaw was reported by The Register. Both are good reminders that self-hosted n8n needs disciplined patch management, not just a docker run on a quiet VM.

Community, scale, and reach

The community numbers we trust most:

• 199,000+ GitHub stars, 59,800 forks (GitHub) — one of the top 50 GitHub projects overall, per n8n’s own home page.
• 40,000+ members on the community forum (n8n vs Zapier) — a number n8n publishes, consistent with the active Discourse community we browsed.
• 1,500+ integrations and 10,000+ community-built workflow templates (GitHub, n8n.io).
• Active meetups on Luma across Ibadan, Dubai, Mumbai, Brisbane, Guadalajara, Leipzig, Karlsruhe, Kathmandu, and more (luma.com/n8n-events).

The community blog cadence is also worth noting. The n8n blog shipped a multi-post series in July 2026 covering LLM security, RAG vs agentic RAG, AI data pipelines, AI agent governance, and open-source workflow automation compared. The editorial focus on production AI is the clearest signal of where the company is investing in 2026.

Honest limitations

n8n is not the right tool for every team, and we would be doing you a disservice if we pretended otherwise.

1. The learning curve is real. Compared to Zapier’s guided flow, n8n expects you to understand concepts like expressions, item linking, and binary data. Non-engineers can absolutely learn it (Huel’s case study makes that point), but the ramp is steeper.
2. Self-hosting is not free in the business sense. You save license cost and take on operational risk. Patch management matters: a real n8n CVE landed in CISA’s known-exploited catalog in March 2026 (The Register).
3. AI features are unevenly distributed. The AI Assistant and some advanced AI features are still in preview on cloud, and self-hosted customers do not yet have access to the AI Assistant. The platform’s docs are explicit that the AI Assistant for self-hosted is “coming soon” with a bring-your-own-key model.
4. Pricing is in EUR and quotas must be modeled. For AI-heavy workloads, the AI credit model can cap the team unexpectedly. There is no top-up option for credits on lower tiers; the answer is to upgrade.
5. Source-available is not OSI open source. The Sustainable Use License is not MIT or Apache. You can run, modify, and self-host freely, but you cannot resell n8n as a hosted service. For most teams this is fine; for some procurement teams it is a non-starter.

What to expect in the next 12 months

Based on the SAP announcement, the public roadmap signals, and the blog cadence, here is what we expect n8n to ship through mid-2027:

1. Deeper SAP Joule Studio integration. The Joule Studio on SAP BTP launch is positioned as a multi-year commercial relationship. Expect pre-built SAP connectors, native BTP service nodes, and tighter SAP identity and access control.
2. MCP everywhere. Expect more MCP tooling — n8n-as-MCP-server patterns for enterprise use cases, plus deeper integration with Anthropic, OpenAI, and Google agent SDKs.
3. Evaluations and AI Assistant maturing on self-hosted. The AI Assistant preview is on cloud today; self-hosted users have been waiting.
4. Continued scaling work. The September 2025 scalability benchmark showed Queue mode hitting 162 requests/second on a C5.4xlarge, 10x the single-instance baseline. Multi-main support and external storage in S3 are still marked as “coming soon” on the pricing matrix.
5. Tighter LLM governance. The July 2026 LLM security guide is the most concrete thing we have seen from any workflow vendor on prompt injection defenses, output validation, and agent constraints. Expect more of that work to land in nodes.

How we would use n8n in 2026

We have run a lot of automation tools, and the honest answer is that the right choice depends on the team. Our quick heuristic:

• Pick n8n if you have engineers who want to see the graph, you need to run workflows on your own infrastructure, you are building AI agents that have to call real systems, or you want to avoid per-step billing on complex workflows.
• Pick Zapier if your automations are owned by non-technical users, your connectors are mainstream SaaS, and you want the lowest possible setup time.
• Pick Make if you want a no-code visual experience but with more complex branching than Zapier offers, and you are willing to think in operations.
• Pick Workato or Tray if you are a large enterprise with strict governance requirements and a budget for a dedicated iPaaS.

If you are choosing n8n, our advice for the first 30 days is straightforward:

1. Start on the free Community Edition with a representative non-production workflow. The npx n8n command gets you to a usable canvas in under five minutes.
2. Pick one workflow that is currently a brittle Python script or a series of Zapier Zaps. Rebuild it in n8n with both deterministic and AI steps. Compare the cost in time and money.
3. If you are regulated, evaluate the SOC 2 and SOC 3 reports, the security page, and the Trust Center. If you need SSO, RBAC, and audit log streaming, the Business or Enterprise plan is where those features live.
4. If you are building AI agents, plan to use Evaluations from day one. The OWASP LLM top 10 is a useful sanity check for what to test.

The bottom line

n8n in 2026 is a fair-code workflow automation platform that has aged into the agentic era without losing its engineering core. SAP’s $5.2B valuation is the loudest signal of where the market thinks orchestration sits in the AI stack (Pulse 2.0, Tech.eu). The quieter signal is the 199,000 GitHub stars, the 1,500+ integrations, and the customers who keep coming back because they can actually debug their own automations.

We will keep watching. If you are shipping with n8n, we want to hear what is working and what is not.

Frequently asked questions

What is n8n used for?

n8n is used to automate multi-step workflows between apps, APIs, and AI models. Teams use it for IT operations, security operations (SOAR), lead management, customer support, RAG chatbots, AI agent orchestration, and internal back-office automation. The product is a node-based visual canvas with custom JavaScript and Python support, 1,500+ integrations, and a self-hostable Community edition (n8n.io).

How much does n8n cost in 2026?

n8n Cloud pricing starts at €20/month for the Starter plan (2,500 executions per month) and goes up to €50/month for Pro and €667/month for Business. Enterprise is custom-priced. Self-hosted Community Edition is free; self-hosted Business and Enterprise are priced separately and add SSO, environments, and governance (n8n pricing).

Is n8n better than Zapier?

It depends on the team. n8n is better for technical teams, AI agent workflows, self-hosted deployments, and predictable per-execution pricing. Zapier is better for non-technical users, simple linear automations, and the broadest possible connector library. We cover the trade-offs in our Zapier AI vs n8n comparison and the n8n vs Zapier page.

Is n8n open source?

n8n is source-available under the Sustainable Use License. The code is visible on GitHub and can be self-hosted for your own automations, but the license does not allow reselling n8n as a hosted service. For strictly OSI-licensed alternatives, the n8n blog’s open-source comparison covers Activepieces, Windmill, Temporal, and Kestra.

What is n8n’s MCP support?

n8n has both a MCP Client Tool node and an MCP Server Trigger. The Client lets your workflow call any MCP-enabled external tool. The Trigger turns any n8n workflow into an MCP server that external agents like Claude or Lovable can call as a tool.

Who founded n8n?

n8n was founded in 2019 in Berlin by Jan Oberhauser. The company is headquartered in Berlin with offices in New York and London, and has raised funding from Sequoia, Felicis, Harpoon, Highland Europe, HV Capital, and SAP (n8n press, EU-Startups).

Is n8n secure enough for enterprise use?

n8n is SOC 2 audited, ships a public SOC 3 report, hosts Cloud data in the EU on Microsoft Azure, encrypts customer data at rest, supports SSO/SAML/LDAP and RBAC on paid plans, and streams audit logs to external SIEM platforms on Enterprise. Self-hosted deployments require disciplined patch management; a critical n8n CVE was added to CISA’s Known Exploited Vulnerabilities catalog in March 2026.

Does n8n work with AI models and agents?

Yes. n8n ships native LangChain integration, an AI Agent node, Evaluations for testing AI workflows, MCP client and server nodes, and support for any chat model provider (OpenAI, Anthropic, Google, Mistral) plus local models via Ollama. AI Assistant features are in preview on cloud and are planned for self-hosted (n8n AI).

What integrations does n8n support?

n8n ships 1,000+ pre-built integrations and 1,500+ total when you include community nodes, plus universal HTTP, GraphQL, and Webhook connectors for anything else (GitHub). The integration library covers databases, communication, productivity, CRM, security, and AI services.

How is n8n different from Make?

n8n charges per full execution and includes unlimited steps and users on every plan. Make charges per operation, with a separate data transfer quota tied to your plan. n8n is also self-hostable, while Make is cloud-only. We cover the trade-offs in Make vs n8n and on the n8n vs Make page.