n8n Review

What n8n is

n8n is a workflow automation platform for technical teams. A visual canvas connects triggers, integrations, transformations, code, databases, and AI components. AI Agent nodes can select tools, use memory, and participate in multi-agent patterns, while normal workflow nodes keep deterministic branches visible. It can run in n8n Cloud or on infrastructure managed by the customer.

This makes n8n more flexible than a simple no-code connector, but flexibility increases operational responsibility. An Agent node should not replace a clear rule. Keep known business logic in explicit nodes and use a model only where variable context requires interpretation. Put tool calls with financial, external, destructive, or permission-changing effects behind approval.

Cloud, Community, Business, and Enterprise

The word “self-hosted” covers several editions. Community Edition is the free standard runtime available from the public repository. Business is a paid self-hosted plan with environments, Git-based version control, SSO, LDAP, scaling, and additional collaboration. Enterprise can be hosted by n8n or self-hosted and adds controls such as external secret stores, log streaming, extended retention, and dedicated support.

At the review date, n8n pricing listed Cloud Starter at €20 per month billed annually for 2,500 executions, and Pro at €50 for 10,000 executions. Business was €667 per month billed annually for 40,000 executions and was self-hosted. Enterprise was custom. The table stated unlimited users, workflows, steps, and integrations across paid plans, with price based on complete workflow executions.

One workflow run counts once regardless of node count, but that does not make AI free. Model tokens, vector databases, external APIs, storage, network traffic, and self-hosted infrastructure remain separate. Concurrency, saved history, projects, insights, environments, and support also vary by plan.

License limits matter

n8n is often described casually as open source. Its Sustainable Use License makes source available and permits many internal business uses, personal use, education, nonprofit activity, and consulting. It restricts uses such as offering n8n itself as a paid competing service or exposing it to customers as part of a commercial product without the appropriate agreement.

Procurement should therefore separate “can inspect and self-host the source” from “may redistribute or monetize it in any way.” A company using n8n internally has a different license question from a SaaS vendor embedding a multi-tenant automation builder. Review the actual intended architecture with counsel rather than relying on a repository badge.

Credentials, tools, and human review

n8n workflows can hold OAuth grants, API keys, database passwords, webhooks, code, MCP connections, and model credentials. Use project separation and distinct service accounts. Give each workflow only the credentials it needs. Avoid importing community workflows without reviewing every node, expression, package, URL, and credential reference.

Human-in-the-loop patterns can pause before sensitive Agent tool calls. The reviewer should see the exact proposed arguments and target. Timeouts and rejections should stop safely. Use idempotency keys and checkpoints so a resumed or retried execution does not duplicate an earlier action.

An AI workflow should cap model steps, retries, runtime, concurrency, and external spend. Test prompt injection from email, web pages, documents, and tool output. A retrieved instruction must not be able to add a new credential or expand an Agent's permitted tool set.

Self-hosting security and operations

n8n's self-hosting documentation says self-hosting requires server, container, scaling, and security knowledge and recommends n8n Cloud for users without that experience. Operators own TLS, reverse proxies, authentication, encryption keys, database backups, queue workers, monitoring, patching, and recovery.

Self-hosting the orchestrator does not keep every payload local. OpenAI, Anthropic, a vector database, a search API, an error tracker, or a connected SaaS may still receive prompts, files, embeddings, credentials, and outputs. Draw the network data path and test deletion in execution history, chat memory, databases, file storage, logs, backups, and downstream apps.

Paid self-hosted licenses periodically communicate usage information to n8n's license service. n8n also describes default telemetry that operators can configure. Teams with strict network controls should document these endpoints and verify the selected edition's behavior.

Who should choose n8n?

n8n is a strong candidate when a technical team wants one canvas for integration logic, code, and AI while retaining a self-hosting path. It is not a shortcut around platform operations. Community Edition is most economical when the team already has infrastructure and security expertise; otherwise engineering time can exceed a cloud subscription.

Compare Make for a managed visual scenario experience, Langflow for an MIT-licensed agent and RAG builder, and Dify for a packaged AI application platform. Use the AI automation and agent tools category to build a common test matrix.