Overview
What does n8n do?
n8n is a visual workflow builder in the Zapier lineage, with two differences that matter for agents: you can run the whole thing on your own infrastructure, and any node in a workflow can be an AI agent with tools attached.
You build on a canvas — triggers on the left, transformations and integrations flowing right. Where a traditional automation would need a rigid if/else tree, you drop in an AI Agent node, give it a model, a system prompt and a set of tools (which are just other n8n nodes), and let it decide which to call. That hybrid of deterministic pipeline plus agentic decision point is what most production automation actually needs.
Because it is source-available and self-hostable, it is the default answer for teams with data residency requirements or an aversion to per-task SaaS pricing. It also means the cost of a high-volume workflow is your server bill plus model tokens, not a per-execution fee.
Features
n8n features
AI Agent node with tools
Attach any n8n integration as a callable tool, so the agent can query Postgres, post to Slack and update HubSpot within one reasoning loop.
400+ native integrations
The broadest connector library of any self-hostable option, plus a generic HTTP node for everything else.
Code nodes
Drop into JavaScript or Python anywhere in a workflow when the visual builder runs out of expressiveness.
Self-hosting with fair-code licence
Run it in Docker on your own box. Free for internal use; the licence restricts reselling it as a competing hosted service.
Native MCP support
Acts as both an MCP client and server, so n8n workflows can expose themselves as tools to other agents.
Execution history and replay
Every run is inspectable node by node, which makes debugging agent behaviour far easier than in black-box products.
Integrations
Integrations and how to connect n8n
What n8n connects to, and the order to set it up in.
Choose hosted or self-hosted
`docker run -it --rm -p 5678:5678 n8nio/n8n` gets you running locally in a minute; n8n Cloud removes the ops burden for a monthly fee.
Add credentials
Connect each service once under Credentials — OAuth for Google and Slack, API keys for most others. Credentials are encrypted and reusable across workflows.
Build the deterministic skeleton first
Trigger, fetch, filter. Resist making the whole workflow agentic; deterministic steps are cheaper and never hallucinate.
Insert an AI Agent node
Give it a model credential, a tight system prompt and only the tools it genuinely needs. Over-provisioning tools is the main cause of erratic behaviour.
Test with pinned data, then activate
Pin sample data to iterate without re-running side effects, then activate the workflow and monitor the execution log for the first few live runs.
Fit
Who is n8n for?
Technical ops and automation engineers
The ideal user. Comfortable with APIs, JSON and a bit of JavaScript, and rewarded with almost unlimited flexibility.
Teams with data residency requirements
Self-hosting means customer data never leaves your infrastructure except for model calls — and you can point those at a local model too.
High-volume automation on a budget
Self-hosted execution is effectively free at the margin, which beats every per-task competitor at scale.
Non-technical business users
Workable for simple flows but the learning curve is real. Expression syntax and JSON handling will block most non-technical users.
Use cases
Best use cases for n8n
Internal ops automation
Onboarding checklists, approval routing, report generation — the long tail of processes nobody wants to build an app for.
Data sync between systems with no native integration
The classic use case, made smarter by letting an agent handle the messy mapping decisions.
Lead enrichment pipelines
Deterministic fetch, agentic research and qualification, deterministic write to CRM. A textbook hybrid workflow.
Scheduled research and reporting
Nightly competitor sweeps or metric digests posted to Slack, at a fraction of the cost of a per-task platform.
Pricing
n8n pricing
Free self-hosted (fair-code); cloud priced per workflow execution. Free tier: Yes — unlimited self-hosted community edition.
Community (self-hosted)
Most commonFree
Full workflow engine on your own server.
- Unlimited workflows and executions
- All integration nodes
- You handle hosting and upgrades
Starter (cloud)
€20/month billed annually
Hosted, for small teams.
- 2,500 executions/month
- Managed hosting
- One execution = one whole workflow run
Pro (cloud)
€50/month billed annually
Higher volume plus collaboration.
- 10,000 executions/month
- Multiple users
- Execution search
Business (cloud)
€667/month billed annually
Large-scale hosted use.
- 40,000 executions/month
- Advanced admin
Enterprise
Custom
Self-hosted with support and governance.
- SSO and LDAP
- Log streaming and audit
- Git-based deployment
Pricing last verified . Vendors change pricing frequently — always confirm on n8n GmbH’s own pricing page before buying.
User sentiment
What other users say about n8n
A summary of publicly available feedback about n8n — not paid testimonials.
n8n is consistently the recommendation when someone asks for automation without vendor lock-in, and the free self-hosted tier drives most of that goodwill. The two persistent criticisms are the learning curve relative to Zapier, and the operational burden of running it yourself once workflows become business-critical.
What users praise
- Self-hosting removes both per-task cost and data residency concerns
- Hybrid deterministic-plus-agentic design is more reliable than fully agentic tools
- Execution history makes debugging genuinely tractable
- Integration coverage rivals commercial competitors
Common complaints
- Noticeably harder to learn than Zapier for non-technical users
- Self-hosting means you own upgrades, backups and uptime
- Cloud execution pricing gets expensive at high volume, undercutting the value argument
- Documentation lags behind the pace of feature releases
We publish individual user reviews only once they have been collected and attributed to a real source. None have been collected for n8n yet. Used it? Send us your review.
Sentiment summarised from: n8n — documentation, n8n — pricing.
Verdict
Is n8n worth it?
Pros
- Genuinely free at unlimited volume if you self-host
- Enormous integration library
- Mix deterministic steps with agent decision points
- No vendor lock-in — workflows are portable JSON
Cons
- Steeper learning curve than no-code competitors
- Self-hosting is real operational work
- Cloud pricing is less compelling than the free tier suggests
- Fair-code licence is not OSI open source
Assessed from vendor documentation, pricing pages and public user sentiment — not a hands-on trial. How we assess agents.
FAQ
n8n frequently asked questions
Is n8n really free?
The self-hosted community edition is free for unlimited workflows and executions under a fair-code licence. You pay for hosting and model tokens. The restriction is on reselling n8n itself as a hosted service.
Do I need to know how to code?
Not strictly, but you will hit expression syntax and JSON handling quickly. Basic JavaScript makes the difference between simple flows and genuinely useful ones.
Can I use my own local model?
Yes. Point the AI Agent node at any OpenAI-compatible endpoint, including a locally hosted model, which keeps the entire pipeline on your infrastructure.
n8n or Zapier Agents?
Zapier if you want the shortest path and have a budget for per-task pricing. n8n if you have technical staff, high volume, or data that cannot leave your infrastructure.
Compare