Getting started with n8n Agents in the cloud or on your own server
· Tutorial
This tutorial explains how to start n8n, including its official path for agents, using only deployment options, plans, and start commands that n8n documented as of 16 November 2026. It does not invent click-by-click console steps that the retrieved docs did not state. Confirm current commands and prices on the official pages before you run or buy anything.
Catalogue page: n8n Agents. See methodology.
n8n is an official workflow automation product. The n8n.ai-agents marketing page describes an AI workflow automation platform and says you can start with an agent, a workflow, or both; add standalone agents or embed them in workflows; watch runs on a visual builder; and choose self-hosting, n8n’s managed cloud, or both. Official docs (“Choose how to use n8n”) say n8n offers two primary deployment options: n8n Cloud (fully managed) and self-hosted (on your own infrastructure). A pricing-page FAQ says workflows can become tools that an agent calls, and that you can build an autonomous agent with the AI Assistant in natural language or manually in the Agent Builder, then do a test run before it goes live.
Steps and surfaces
Official docs present two decisions: cloud or self-hosted, then which plan or edition. For a first start, the same “Choose how to use n8n” page says: “Start with n8n Cloud for instant access with no installation required,” and it points to a free trial. The Cloud docs overview (“Use n8n Cloud”) says you manage a trial, plan, and instance in the n8n Cloud console, and it points to a “Try free then choose a plan” guide for starting a trial and comparing Starter, Pro, and Enterprise. This tutorial does not invent the console clicks; use that official Cloud guide for the current sign-up steps.
If you want to self-host, the official “Host n8n” page says you can run n8n on your own infrastructure, on-premises, or in a private cloud, using Docker Compose, a one-line setup, or other methods. For a quick local setup the same page publishes this command, with the instruction “Don’t want to read the docs? Run this:”
curl -fsSL https://get.n8n.io | sh
The docs say this requires Docker on Linux or macOS (or WSL on Windows), “sets up n8n locally in one step,” and they point to a one-line setup guide for what the script does. Re-check that page before you run the command, because install instructions change. The same hosting page compares other official methods: Docker Compose (recommended there for production, requiring Docker and Docker Compose); cloud-provider guides (including AWS, Azure, Google Cloud, DigitalOcean, Hetzner, Heroku, and OpenShift); and npm, which the docs say is “Best for: Local development or testing” and “deprecated from n8n 3.0,” with a recommendation to use Docker Compose or the one-line setup instead.
All self-hosted installations, the hosting docs say, use the same core product. Without a license key, n8n runs as the free Community edition; a Business or Enterprise license key enables those editions. The “Choose how to use n8n” page adds a free Registered Community option: Community plus folders, debug in editor, and custom execution data, unlocked by registering an email at Settings > Plan & usage > Unlock.
After n8n is running, official marketing and pricing FAQs describe how to add an agent: start with a ready-made workflow or from scratch; build a standalone agent or embed an agent in a workflow; use the AI Assistant (recommended in the FAQ) or the Agent Builder; run a test before publishing. A pricing FAQ says session memory is the default, and that cross-session (episodic) memory is opt-in and needs an OpenAI credential. The same FAQ says one turn with an agent is one execution, that sub-agent and workflow-tool calls do not add extra executions, and that agents share the execution quota with workflows. It also says external systems cannot yet call n8n agents from external agents and MCP servers; that capability is described as on the roadmap. Stay inside those official limits; this tutorial does not invent an MCP callback procedure.
Plan and pricing notes
As of 16 November 2026, the official pricing page (https://n8n.io/pricing) said: “All plans include unlimited users and workflows and every integration. Pricing is based on monthly workflow executions, regardless of complexity.” A FAQ defines an execution as a single run of an entire workflow, regardless of how many steps it contains or how much data it processes.
For n8n Cloud, the pricing page listed Starter (“Build and run your first workflows and agents”) at 20€ per month, billed annually, with no credit card required, 1 shared project, 5 concurrent executions, and 1,600 Assistant credits per month. Pro (“For solo builders and small teams running automations in production”) was listed at 50€ per month, billed annually, also with no credit card required, including everything in Starter plus 3 shared projects, up to 50 concurrent executions, up to 9,600 Assistant credits per month, and workflow history. Business was listed at 667€ per month, billed annually, with the retrieved card text “For companies with < 100 employees needing collaboration and scale,” and a feature line “Self-hosted only” also appeared on the same pricing page; use the official pricing table for the current Cloud-versus-self-hosted placement of Business. Enterprise is listed as a higher tier with, among other items, unlimited shared projects and 200+ concurrent executions. The pricing page also mentions a Startup plan: under 20 employees may qualify for 50% off Business. A Community Edition is described as “A standard, self-hosted version of n8n is available on GitHub.”
The official Cloud trial FAQ said a free trial of n8n Cloud includes Pro-level features (including Global Variables, Insights with a dashboard and 14-day history, 5-day workflow history, advanced execution filters, unlimited evaluations, folders, and Assistant), with trial limits of 1,000 executions, 5 concurrent executions, a 180-second execution timeout, and 1 team project, plus 800 Assistant credits and 100 AI credits. Admin roles, the project Editor role, Named Versions, Workflow Diffs, and the n8n public API are listed as not included in that trial. No credit card is required for the n8n Cloud trial; a credit card is required for the Business plan free trial, according to the same FAQ. “Choose how to use n8n” describes the Cloud free trial as 14 days of Pro features.
The official pages retrieved on 16 November 2026 showed annual euro prices for Starter, Pro, and Business. Monthly list prices were not clearly labelled in the retrieved HTML next to those plan names, so they are not quoted here; see the current pricing page. Confirm current prices before you enrol.
Which use case this fits
This tutorial fits someone who has already decided to try n8n and needs the official start path on n8n Cloud or on their own machine, including the documented way to add an agent and how executions are counted. It fits both a first cloud trial (no install, official 14-day Pro-level trial) and a self-hosted Community edition start (free, with the official one-line command or Docker Compose). It is not a comparison of n8n with other automation vendors, and it is not a guide to a specific third-party LLM vendor’s API, even though official AI docs say n8n can connect providers such as OpenAI, Anthropic, and Google.
Closing advice
Start from the official “Choose how to use n8n” page. If you want the fewest setup steps, use n8n Cloud and the official “Try free then choose a plan” guide. If you want to run the Community edition yourself, use the documented one-line setup or Docker Compose, and treat the npm path as deprecated from n8n 3.0 unless the current docs still recommend it for a short local test. After the instance is up, add an agent with the official Assistant or Agent Builder, run a test, and remember that official pricing counts one agent turn as one execution against the same quota as workflows. Then re-check the live docs and pricing page, because commands, editions, and allowances change. Related pages: n8n Agents and methodology.
Sources
Retrieved on 16 November 2026 (Asia/Shanghai):
- n8n, “AI Workflow Automation Platform - n8n,” https://n8n.io/ai-agents/. Product framing for workflows and agents, start options, and deployment choice (cloud, self-host, or both).
- n8n, “n8n Plans and Pricing - n8n.io,” https://n8n.io/pricing. Plan names, annual euro list prices, execution definition, trial FAQ, assistant-credit allowances, and agent execution rules.
- n8n Docs, “Choose how to use n8n,” https://docs.n8n.io/choose-n8n/ (canonical URL retrieved: https://docs.n8n.io/choose-how-to-use-n8n). Cloud versus self-hosted decision, plan and edition list, 14-day Cloud trial, and Registered Community unlock path.
- n8n Docs, “Host n8n,” https://docs.n8n.io/hosting/ (canonical URL retrieved: https://docs.n8n.io/deploy/host-n8n). Official one-line install command, Docker and npm notes, and Community versus licensed editions.
- n8n Docs, “Use n8n Cloud,” https://docs.n8n.io/manage-cloud/overview/ (canonical URL retrieved: https://docs.n8n.io/deploy/use-n8n-cloud). Console, trial, and plan-management overview.
- n8n Docs, “Integrate AI,” https://docs.n8n.io/advanced-ai/ (canonical URL retrieved: https://docs.n8n.io/build/integrate-ai). Official statement that n8n can connect LLM providers such as OpenAI, Anthropic, and Google, and can add tools and memory.