How much does an n8n workflow cost to build?

A simple workflow joining two or three systems, for example a webhook into Airtable with an email notification, costs $70 to $400 and takes one to three days. A workflow with AI steps, a CRM sync and real error handling costs $500 to $1,500. Larger automation programmes with several connected workflows are commonly quoted at $3,000 to $6,000.

Is n8n cheaper than Zapier or Make?

Usually yes at volume, because n8n prices by workflow execution rather than per task, and you can self-host it. A workflow that fires a thousand tasks a month is dramatically cheaper on n8n. Below a few hundred tasks a month the difference is small enough that the better tool is whichever one your team can maintain.

Should I use n8n Cloud or self-host it?

Use n8n Cloud unless you have a specific reason not to. It starts around $20 to $50 a month and removes upgrades, backups and uptime from your plate. Self-host when you have data that cannot leave your infrastructure, when execution volume makes Cloud pricing worse than a server, or when you need custom community nodes. Self-hosting costs roughly $10 to $40 a month in server fees plus the maintenance nobody budgets for.

What makes an n8n automation expensive?

The number of separate systems it touches, and whether failure is handled. Each integration brings its own authentication, rate limits and error shapes. A workflow that retries sensibly, avoids duplicate actions, logs what it did and alerts a human when it gives up costs roughly twice a workflow that assumes everything succeeds, and is worth it the first time an API has a bad morning.

When should I not use n8n?

When the logic is the product rather than the plumbing, when you need per-request latency under a second, or when a workflow has grown past roughly forty nodes with branching that nobody can follow. At that point you are writing software in a diagram, and actual code with tests is cheaper to change. n8n is excellent at moving data between systems on a schedule or a trigger, and that is what it should stay doing.

Where can I hire an n8n developer?

Search for an n8n developer or an automation agency, and ask to see the inside of a workflow they built rather than a demo of it running. The screening question that separates professionals from hobbyists is how they handle API rate limits, webhook failures and retries. Anyone who cannot answer that concretely has only built workflows that have not broken yet.

Automation6 October 2026 · 7 min read

What Does an n8n Automation Cost to Build?

The short answer

A single n8n workflow connecting two or three systems costs $70 to $400 and takes one to three days. A multi-step workflow with AI steps, a CRM and proper error handling runs $500 to $1,500. Ongoing work for a business running automations day to day is usually $20 to $35 an hour or a monthly retainer. The price is set by how many systems the workflow touches and how carefully it handles failure, not by how many steps it has.

n8n jobs get posted at $40 and at $6,000, often for work that sounds similar in the brief. The difference is almost never the number of steps. It is how many separate systems the workflow touches, and whether anyone paid for it to handle a bad day.

What n8n workflows cost

Scope	Price	Timeline
One trigger, two or three systems, happy path only	$70 to $400	1 to 3 days
Multi-step with an AI step, a CRM sync, real error handling	$500 to $1,500	3 days to 2 weeks
A set of connected workflows running a business process	$3,000 to $6,000	3 to 6 weeks
Ongoing maintenance and new workflows	$20 to $35 per hour, or a monthly retainer	Ongoing

Real examples from recent postings, to calibrate: a webhook into Airtable plus an email receipt at $250. Google Drive to an AI captioning step to a social scheduler at $1,000. AI lead qualification writing into HubSpot at $100. A workflow publishing content to an S3 site at $40.

The $40 and $100 jobs are real, and they are a reasonable price for one trigger, one transform and one destination, assuming nothing is expected to go wrong.

What actually drives the price

The number of systems, not the number of nodes. Each system you connect brings its own authentication, its own rate limits, its own pagination and its own way of failing. A fifteen-node workflow across two systems is cheaper than a six-node workflow across five.

Whether failure is handled. This is the real split between a $200 workflow and a $900 one. More on it below.

Whether the data is clean. "Pull the leads from the spreadsheet" costs very little when the spreadsheet is consistent and quite a lot when three people have been typing dates in different formats since 2022.

Self-hosted or cloud. Self-hosting adds setup, and it adds someone being responsible for upgrades at a bad moment.

n8n Cloud or self-hosted

Use Cloud unless you have a specific reason not to. It runs about $20 to $50 a month at small volume and takes upgrades, backups and uptime off your plate.

Self-host when one of these is true:

Data genuinely cannot leave your infrastructure, for example health or financial records
Execution volume makes Cloud pricing worse than a server, which happens later than most people assume
You need community nodes that Cloud does not offer

Self-hosting runs roughly $10 to $40 a month on a small VPS, plus maintenance nobody puts in the budget. A self-hosted n8n that nobody has upgraded in a year is a liability, not a saving.

Versus Zapier and Make

n8n generally wins on cost at volume because it prices by workflow execution rather than per task. One execution that touches a hundred records is one execution. On Zapier that can be a hundred tasks.

Below a few hundred tasks a month the price difference is small enough that the right tool is the one your team can actually maintain. Zapier is easier for a non-technical person to pick up. n8n gives you branching, code steps and self-hosting. If nobody at your company will ever open the workflow, that advantage is theoretical.

The part that separates a workflow that survives

Most quotes cover the happy path. The ones worth paying for cover these five, and this is also the checklist to interview against:

Retries with backoff. APIs fail transiently. A workflow that gives up on the first 500 will lose data on an ordinary Tuesday. Retries must back off, or you turn a brief outage into a rate-limit ban.

Rate limit handling. Every API has one. The workflow needs to know what it is and pace itself, not discover it in production at the worst moment.

Duplicate prevention. Webhooks get delivered twice. Retries re-run steps. Without an idempotency check, a retried workflow sends the same invoice twice. This is the failure that costs money and trust rather than time.

Failure alerting. A workflow that stops silently is worse than one that never ran, because you believe it is working. There should be an error branch that notifies a human, with enough context to see what broke.

Execution logging. When someone asks why a record did not sync three weeks ago, you need the run. n8n keeps execution history, but check the retention on your plan, because the default is shorter than the gap between a problem and the complaint about it.

A quote that does not mention any of these is a quote for the demo, not the system.

When n8n is the wrong tool

Three cases where writing actual code is cheaper:

The logic is the product. If the valuable part is a pricing engine or a matching algorithm, you will spend more fighting the canvas than writing it.
You need sub-second responses. n8n is built for triggers and schedules, not for sitting in a user-facing request path.
The workflow passed about forty nodes with branching nobody can follow. At that point you are writing software in a diagram, with no tests and no code review. Pull the complex part into a small service and let n8n call it.

The last one creeps up. A workflow that was clear at ten nodes becomes unmaintainable at fifty, and because it is still technically working, nobody schedules the rewrite until it breaks.

Hiring an n8n developer

What to search for. "n8n developer", "automation developer", "automation agency". Specify n8n rather than "Zapier expert" if you have chosen it, because the error-handling patterns are what you are buying and they do not transfer directly.

The one screening question that works: how do you handle API rate limits, webhook failures and retries in a complex workflow? Someone who has run automations in production answers this immediately and in specifics. Someone who has only built tutorials talks generally about error handling. One agency we saw posting for this asks exactly that question and auto-declines proposals that skip it, which tells you how reliably it separates people.

Ask to see inside a workflow, not a demo of one running. A screen recording of the canvas showing the error branches is worth more than a video of a successful run.

What to get in writing: the workflows exported as JSON into your own repository, credentials in your own accounts, a documented list of what each workflow does and what it touches, and who to call when one breaks.

This is AI agents and automation work, and the backend engineering side covers the point where a workflow should become a service. The AI sales and outbound automation and autonomous multi-agent operations case studies both show the shape at the larger end.

If your automation includes an AI step, how to tell if your AI agent is actually working covers the part that silently goes wrong.

Common questions

How much does an n8n workflow cost to build?: A simple workflow joining two or three systems, for example a webhook into Airtable with an email notification, costs $70 to $400 and takes one to three days. A workflow with AI steps, a CRM sync and real error handling costs $500 to $1,500. Larger automation programmes with several connected workflows are commonly quoted at $3,000 to $6,000.
Is n8n cheaper than Zapier or Make?: Usually yes at volume, because n8n prices by workflow execution rather than per task, and you can self-host it. A workflow that fires a thousand tasks a month is dramatically cheaper on n8n. Below a few hundred tasks a month the difference is small enough that the better tool is whichever one your team can maintain.
Should I use n8n Cloud or self-host it?: Use n8n Cloud unless you have a specific reason not to. It starts around $20 to $50 a month and removes upgrades, backups and uptime from your plate. Self-host when you have data that cannot leave your infrastructure, when execution volume makes Cloud pricing worse than a server, or when you need custom community nodes. Self-hosting costs roughly $10 to $40 a month in server fees plus the maintenance nobody budgets for.
What makes an n8n automation expensive?: The number of separate systems it touches, and whether failure is handled. Each integration brings its own authentication, rate limits and error shapes. A workflow that retries sensibly, avoids duplicate actions, logs what it did and alerts a human when it gives up costs roughly twice a workflow that assumes everything succeeds, and is worth it the first time an API has a bad morning.
When should I not use n8n?: When the logic is the product rather than the plumbing, when you need per-request latency under a second, or when a workflow has grown past roughly forty nodes with branching that nobody can follow. At that point you are writing software in a diagram, and actual code with tests is cheaper to change. n8n is excellent at moving data between systems on a schedule or a trigger, and that is what it should stay doing.
Where can I hire an n8n developer?: Search for an n8n developer or an automation agency, and ask to see the inside of a workflow they built rather than a demo of it running. The screening question that separates professionals from hobbyists is how they handle API rate limits, webhook failures and retries. Anyone who cannot answer that concretely has only built workflows that have not broken yet.

Work we do on this

AI Agents & Automation AI Sales & Outbound Automation Autonomous Multi-Agent Operations

Want this looked at properly?

Tell us what you're working with. We'll give you a straight answer in a 30-minute call, free.