--- title: "n8n Cloud vs. Your Own n8n Instance: Real Cost Breakdown" description: "n8n Cloud charges per execution and requires your own AI API key. Cost breakdown vs. owning your n8n instance with AI credits included from $19.99/mo." canonical: https://agentroost.app/en/blog/n8n-cloud-vs-self-hosted-cost-breakdown date: 2026-05-20T20:00:00Z --- [Canonical URL](https://agentroost.app/en/blog/n8n-cloud-vs-self-hosted-cost-breakdown) If you've priced out n8n Cloud recently and felt the sticker shock creep up as you added AI nodes, you're not alone. The execution-based billing model that n8n Cloud uses is sensible at small scale — but it has two hidden costs most comparisons skip: the **execution meter** and the **AI API key you still have to fund yourself**. This post breaks both down line by line, compares them against running your own n8n instance with AI credits already included, and helps you decide which option actually costs less for your workload. --- ## What n8n Cloud actually charges for n8n Cloud uses a **tiered, execution-based pricing model**. Every time a workflow runs, it burns executions from your monthly quota. Run a webhook-triggered workflow 500 times a day? That's ~15,000 executions per month — and depending on your plan, you may hit the ceiling before the month ends. Key things to understand about the n8n Cloud model: - **Executions are metered.** Each workflow run counts, regardless of whether it did something useful or failed mid-way. - **More executions = higher plan = more cost.** Scaling a chatbot or a monitoring loop to run every 5 minutes quickly outgrows the lower tiers. - **AI nodes are not included.** When you drop an AI Agent node, an Embeddings node, or any LLM-powered step into a workflow, n8n calls whichever provider you've connected. That means you need an OpenAI API key (or Anthropic, Groq, etc.), billed separately, from your own account. n8n Cloud provides the orchestration; you pay the model bill on top. So a realistic monthly cost for a moderately active n8n Cloud user running AI workflows might look like: | Line item | Approx. cost | |---|---| | n8n Cloud plan (Starter/Pro) | $20–$50+ | | OpenAI / Anthropic API usage | $10–$50+ (usage-dependent) | | Execution overages (if any) | $0–$20+ | | **Total** | **$30–$120+** | These aren't invented numbers — they're the math you get when you run even a modest set of AI-powered automations at any real frequency. --- ## What "your own n8n instance" means Before going further: **there are two ways to own your n8n instance**, and they have very different cost/effort profiles. ### Option A: True self-hosting (you run it) You spin up a VPS (€4–€10/mo on Hetzner, DigitalOcean, etc.), install Docker, configure n8n with environment variables, set up a reverse proxy for HTTPS, manage SSL renewals, watch disk usage as execution logs accumulate, handle upgrades, and make sure it stays running. When it goes down at 2am, you fix it. Cost is low — maybe €6–€8/mo for a small server — but the hidden cost is **time**. First setup easily takes 3–4 hours. Ongoing maintenance is unpredictable. You still need to wire in your own AI API keys and manage the billing there. ### Option B: Your own single-tenant instance, DevOps done for you You get a private n8n instance — your login, your data, your workflows at `https://.agentroost.app` — without touching a server. The infrastructure is handled for you. **The key difference from n8n Cloud:** it's a flat monthly price, executions are not metered, and **AI credits are already included**. You open the AI Agent node and it works. No API key setup, no separate AI bill. --- ## Side-by-side comparison | | n8n Cloud | True self-host | AgentRoost (your own instance) | |---|---|---|---| | **Setup time** | ~5 min | 3–4 hours | ~2 minutes | | **Execution metering** | Yes (quota-based) | No | No | | **AI API key required** | Yes (BYOK) | Yes (BYOK) | No — credits included | | **AI model bill** | Yours, separate | Yours, separate | Included in subscription | | **You own the instance** | No (shared infra) | Yes | Yes (single-tenant) | | **DevOps burden** | None | High | None | | **Starting price** | ~$20+/mo + AI bill | ~€6/mo + AI bill + time | $19.99/mo all-in | | **Execution logs / history** | Limited by plan | Limited by disk | Instance-level | | **Upgrades** | Automatic | Manual | Handled | | **Cancel anytime** | Yes | N/A | Yes, 14-day money-back | --- ## The AI credit math that changes everything This is the comparison most blog posts skip. If you run any meaningful AI automation — a content pipeline that calls an LLM to summarize or categorize, a support bot, a lead-enrichment workflow — you're paying for LLM tokens on top of your n8n bill. That spend is invisible when you're shopping n8n Cloud plans, but it shows up every month in your OpenAI or Anthropic invoice. On AgentRoost, the AI credits are bundled in. The AI Agent node, LLM nodes, and embedding nodes in your n8n instance are already connected. You don't create an API key, you don't add a payment method to a model provider, you don't watch a separate dashboard for usage spikes. The $19.99/mo base plan includes a working AI-wired n8n instance — full stop. > **The honest comparison:** n8n Cloud Starter (~$20) + a modest OpenAI usage bill (~$15) puts you at $35+/mo for metered, AI-capable automation where you don't own the instance. AgentRoost at $19.99/mo gives you a single-tenant instance you fully own, no execution metering, AI included. If your workload grows, Plus and Pro tiers on AgentRoost add more compute and more included credits — still a flat price, still no separate AI bill. --- ## Who should pick what **n8n Cloud makes sense if:** - You're evaluating n8n for the first time and want the fastest possible start with no commitment. - You run very few workflows and execution counts stay comfortably within the lower tiers. - AI nodes aren't part of your setup yet. **True self-hosting makes sense if:** - You're a developer comfortable with Docker and Linux. - You want maximum control over storage, network, and upgrade timing. - You're running very high execution volumes where flat-rate infrastructure beats per-execution pricing. **Your own instance on AgentRoost makes sense if:** - You want to own your n8n instance without owning the infrastructure. - You're building AI-powered workflows and don't want a surprise AI API bill each month. - You value a flat, predictable price and zero server maintenance. --- ## How to run your own n8n instance on AgentRoost The actual flow takes about two minutes: 1. Sign up at [agentroost.app](/en/agents/n8n) — email/password or Google/Microsoft/Discord. 2. Pick the **n8n** framework, give your instance a name. 3. Your private n8n editor opens at `https://.agentroost.app`. 4. Drop in any workflow. Add an **AI Agent** node — it's already wired to included credits. No API key prompt, no setup step. 5. Build. Webhooks get a real public HTTPS URL immediately. That's it. There's no server to configure, no SSL certificate to provision, no Docker Compose file to write. [Compare plans](/en/pricing) — the base tier starts at $19.99/mo all-in, with a 14-day money-back guarantee. You can also explore the [n8n framework page](/en/agents/n8n) to see what's included. --- ## Bottom line n8n Cloud's execution meter and bring-your-own-AI-key model mean your real cost is always higher than the plan price shown at checkout. For anyone building AI-powered automations, the gap between "plan price" and "actual monthly spend" can easily double. Your own single-tenant instance with AI credits already included isn't just cheaper on paper — it removes an entire category of billing complexity. You know what you pay, month to month, before you build a single workflow.