What we build with n8n
n8n development with us covers four kinds of work: designing and building n8n workflows for real business processes, setting up and running self-hosted n8n, writing custom n8n nodes for APIs that have no ready node, and adding AI steps and AI agents to n8n workflows. We are a software development company, so the same senior engineers who write our back ends also build and review the workflows, and they work with AI coding agents to move faster on the routine parts.
n8n is a good fit for technical teams: it supports branching, loops, JavaScript and Python code steps, sub-workflows and error workflows, it bills per workflow run rather than per step on its cloud plans, and it can run on your own server. Those strengths are also why n8n setups tend to grow into something only one person understands. Our job is to build n8n workflows that the next engineer can read, test and change.
Typical n8n workflows we build
These are the workflows companies ask us for most. Each one usually combines a trigger, a few lookups, a decision and an update in another system.
Lead capture and CRM. A form, ad lead, email or Telegram message arrives, n8n checks for duplicates in the CRM, enriches the contact, assigns an owner by rules and posts a notification to Slack or Telegram. Leads that match no rule go to a review list instead of disappearing.
Orders and invoices. A paid order creates or updates the customer, the invoice and the shipment, and sends the confirmation. Payment and delivery status callbacks come back through webhooks and move the order forward.
Document processing. Invoices, purchase orders or application forms arrive by email, an AI step with OpenAI or Claude extracts the fields into a fixed schema, rules check the result against your records, and anything uncertain goes to a person before it reaches the accounting system.
Notifications and reports. Status changes, overdue tasks and daily or weekly summaries are posted to Telegram, Slack or email, built from Google Sheets, your database or your CRM.
Data sync. Records created in one system appear correctly in others, with mapping, validation and a log of what changed.
Integrations for Ukrainian businesses
For companies in Ukraine, the integration list usually includes Nova Poshta, monobank or PrivatBank, Prom.ua or Rozetka, KeyCRM or SalesDrive, Google Sheets and Telegram. Here is how each connects to n8n, based on n8n's integration directory as we checked it on 19.11.2026:
| Service | How it connects to n8n |
|---|---|
| Telegram, Google Sheets, Slack, Gmail | Built-in n8n nodes |
| OpenAI, Anthropic Claude | Built-in n8n nodes, plus the AI Agent node |
| Nova Poshta | HTTP Request node against the Nova Poshta API; no built-in node found |
| monobank, PrivatBank | HTTP Request node and webhooks against the banks' business APIs; no built-in node found |
| Prom.ua, Rozetka | HTTP Request node against the marketplace seller APIs; no built-in node found |
| KeyCRM, SalesDrive | HTTP Request node and webhooks against their open APIs; no built-in node found |
The HTTP route works well, but it spreads authentication and request details across many workflows. When the same API is used in several places, we write a custom n8n node for it, so tokens, pagination, rate limits and the common operations live in one tested package.
Self-hosted n8n: on your servers or ours
Self-hosting is the main reason many teams choose n8n. Data and credentials stay inside your infrastructure, and there is no per-run fee, only the server. We set up self-hosted n8n in one of two ways:
- On your infrastructure. A server or container in your cloud account or data center. You own the machine and the data; we handle setup and, if you want, ongoing maintenance.
- On a server we manage. Useful when you have no infrastructure team. We run it on a dedicated server for your company, never shared with other clients, and hand it over whenever you ask.
A production setup includes PostgreSQL instead of the default SQLite, queue mode with separate workers when volume needs it, HTTPS, an encryption key stored safely, daily backups, uptime and error alerts, and a tested update procedure. Workflows are exported as JSON to a Git repository, so changes have history and can be reviewed. Self-hosting also means someone is responsible for updates and security patches; if no one should be, n8n Cloud is the more honest choice and we build on it just as well.
Custom n8n nodes
A custom node is a small TypeScript package that adds a new integration or operation to n8n's editor. We write custom nodes when:
- an API you use daily has no node and the HTTP Request setup is repeated across many workflows;
- a step needs logic that is awkward in a code node, such as signing requests, handling files or long-running operations;
- you want your own internal system to appear in n8n as a first-class node for your operations team.
Custom nodes get tests, a version number and installation notes, and they install on self-hosted n8n.
AI agents and AI steps in n8n
n8n ships an AI Agent node and nodes for OpenAI and Anthropic models, so a workflow can classify messages, extract data from documents, draft replies or let a model pick one of several tools. We use these where the input is unstructured and keep everything else as plain rules, because rules are cheaper and predictable.
What we add is control: every model output is parsed against a fixed schema, prompts are kept in version control rather than only in a node's text box, a set of real past examples is re-run before a prompt change goes live, each workflow has a monthly cost limit, and low-confidence results go to a person. When an agent needs many tools, long chains of decisions or strict permissions, we build it as code; see our AI agent development page.
n8n pricing: what the tool itself costs
n8n's own fees are separate from our work. As listed on n8n.io/pricing when we checked it on 19.11.2026, with prices for annual billing:
| n8n plan | Price (billed annually) | Executions per month | Hosting |
|---|---|---|---|
| Community edition | No license fee | Limited by your server | Self-hosted, you pay for the server |
| Starter | 20 EUR per month | 2,500 | n8n Cloud |
| Pro | 50 EUR per month | 10,000 | n8n Cloud |
| Business | 667 EUR per month | 40,000 | Self-hosted |
| Enterprise | On request | Custom | Cloud or self-hosted |
n8n counts one execution per full workflow run, regardless of how many steps it has. Plans and prices change often, so check the page before you decide.
When n8n is the right choice and when to move to code
n8n is the right choice when workflows connect several systems with real logic, when volume is too high for per-step billing, when data must stay on your servers, and when you have, or hire us as, someone technical to own it.
Move a workflow to code when it becomes core to how you make money: when it writes directly into your own database or a legacy system, needs approvals, roles and an audit trail, needs automated tests before every change, or when failures reach customers before you notice them. In that case we build a small service with a queue, retries, logs and a review screen, and n8n can keep calling it for the edges. Our custom software development team builds those services, and our guide n8n vs Zapier vs custom code lists the signs in detail. If your workflows run on Make or Zapier today, see our Make automation and Zapier automation pages.
How we work on n8n projects
- Review (about 1 week). We map the process or read your existing n8n workflows, measure volume and failures, and decide what stays in n8n, what needs a custom node and what should be code. You get a fixed quote.
- Set up. Production n8n instance, credentials, Git export, monitoring and backups, or a review of your current instance.
- Build. Workflows with error workflows, retries and alerts, sub-workflows for shared logic, and test runs on real sample data.
- Shadow run. Where a workflow replaces manual work or an older automation, both run side by side until the results match.
- Handover. Documentation of each workflow, a map of how they call each other, and access for your team. We can stay on for maintenance.
Our minimum project size is $10,000. We do not hold an official n8n partner status; what we bring is a senior engineering team that also writes the code around n8n. We work from Ukraine with companies in the US and Europe. Broader automation work, including processes that do not involve n8n at all, is on our AI automation services page.
Next step
Send us a short description of the process you want automated, or an export of the n8n workflows that cause trouble today. We will tell you what should stay in n8n, what should move to code and what a fixed quote would look like. Contact us to book a call.