Przemysław Zalewski

For agencies

A senior under your brand, for the parts n8n cannot do

You sell automation and AI to small businesses. Then a client needs a custom integration, a real app, a deploy that survives, or an AI feature beyond one prompt. I build that part under your brand, and your client keeps talking to you.

See price and start date →

$900 to $2,700 a month excl. VAT: 20 to 60 hours at the base rate, $45/h. Details in the configurator →

Przemysław Zalewski

Sound familiar?

Problems AI agencies bring me in 2026

Does our client’s self-hosted n8n have a critical vulnerability?

On 8 January 2026 n8n disclosed CVE‑2026‑21858, rated CVSS 10: on versions from 1.65.0 before 1.121.0, certain form workflows let an unauthenticated attacker read files on the server.

I upgrade the client’s instance in waves you can revert, put it behind backups and monitoring, and leave a runbook, so your team knows what runs where when something breaks at night. I have run client automations in Make.com and Zapier myself: in 2025 I routed newsletter and contact forms from about ten client sites through a Make webhook with bot checks into one team channel, still running on Make’s free plan a year and a half later. I also put a scheduled trigger behind an authenticated webhook with a lock against parallel runs.

OpenAI Assistants API shut down and the client’s bot broke

OpenAI removed the Assistants API on 26 August 2026, a year after announcing it; the replacement is the Responses and Conversations APIs (OpenAI deprecations).

I move the bot to the current API, run the same test questions before and after, and add a cost cap. The client gets a bot checked against its old answers, with a known maximum bill. In June 2026 a client’s AI email assistant hit its model provider’s prepaid limit, so I made it reply with a plain unavailability notice instead of bouncing, alert on quota errors, and added a daily end‑to‑end test that emails it and checks the answer.

Client app built with Lovable exposes its database

UpGuard reported on 14 October 2026 that it had found 16,326 Supabase databases with publicly readable tables on sites created by AI coding agents, over half with signs of personal data. Supabase is the database Lovable apps typically run on.

I turn the prototype into a real app under your brand: logins, roles and database rules covered by tests, deployed with backups and monitoring. In 2022 I found a public endpoint writing straight to a client’s database, traced its only consumer and moved that consumer to a proper data migration, so the database could be taken off the public internet without analytics losing data.

Facing one of these? Book a free 30‑min call →

Scope

Where n8n and Make usually end

A real app around the workflow

A client panel, logins, roles and data the client owns, not a spreadsheet behind a webhook.

Integrations without a ready node

Polish ERPs, accounting, warehouse and booking systems with odd APIs, files or no API at all.

AI beyond one prompt

Search over the client’s documents, evaluation of answers, cost caps, model choice.

Deploys that survive

Hosting, backups, alerts and error tracking that report a failure before your client does, and a runbook your team can follow at night without me.

Process

How it works

  1. A capacity package

    20 to 60 hours a month at a fixed rate, planned month by month.

  2. Your brand

    NDA, your repositories or mine, no contact with your client unless you want it.

  3. Weekly status

    A short written status you can forward to the client as is.

  4. Handover

    Every project ends with a runbook, so your team can run it without me.

What I do not do here

  • Sell to your clients or take them over.
  • Security audits.
  • Projects without a clear owner on your side.

Track record

Proof

15+ years shipping software, 9 in TypeScript, React and Node.js. Three recent AI cases below: an AI email client through Google’s CASA Tier 2 with pentest findings fixed, an MCP server shipped to external users, and coding agents held to rules I enforce in review. Client names under NDA stay private.

Patriot Mail: an AI email client through CASA Tier 2 and onto the App Store

The problem
A greenfield AI‑native email client for web, iOS and macOS, built at Sanddev, needed Google’s CASA Tier 2 security assessment before it could read Gmail, and it had to work for older people.
What I did
I was the principal engineer at Sanddev. I wrote the delivery plan, shaped the product vision and treated prompt injection through untrusted email as a design constraint before any model touched a mailbox. CASA and OWASP ASVS controls became a Playwright suite that runs on every pull request. WCAG 2.2 AA sits behind an automated axe gate.
Result
I led CASA Tier 2 to a pass in January 2026, with third-party pentest findings fixed, and led the yearly reassessment, validated in October 2026; the questionnaire turned into a check on every PR. Live on the App Store on iOS and macOS since April 2026. Text scaling to 200% and a calm‑email mode shipped for older users. At that reassessment five items came back for missing evidence; I fixed the code first, then resubmitted updated evidence, and all five were accepted on 27 October.

CASA Tier 2 iOS + macOS WCAG 2.2 AA Open ↗

See also: SaaS & MVP · Mobile & desktop apps

Offer: From an AI‑built MVP to a product people pay for

An MCP server from first merge to first external users

The problem
Users of Sandtime.io, the time tracker I built at Sanddev, used by 4,000+ freelancers and teams, wanted to log time from their AI assistants. The API existed, the agent surface did not.
What I did
I set the tool surface, removed destructive tools, routed every operation through the existing REST controllers so an agent gets no wider access than a person, and instrumented adoption per client in PostHog from day one.
Result
First external users. A bot-protection rule that rejected MCP clients by user agent was found and removed the same day because the dashboard showed it.

MCP OAuth adoption dashboard

See also: AI & agents · PostHog & product analytics

Offer: Your company systems inside Claude and ChatGPT (MCP)

An AI coding agent as a team member, under enforced rules

The problem
Coding agents write code faster than a team can review it. A team of four wanted the speed without a slop swamp on main.
What I did
I authored the rule set shared by Claude Code and Codex: mechanical checks over written rules, lint and test gates for security, API surface and CSP, agents working on main with hooks, every PR reviewed by a human, AI‑assisted content disclosed as such.
Result
In 2026 about two thirds of merged pull requests across two products were built with coding agents, and nearly all went through my review or merge. I merged 260+ PRs of my own and reviewed 300+ of other engineers’. My commits across those two products went from under 1,000 in 2025 to 7,500+ by October 2026.

~2/3 PRs with agents 300+ reviews 7,500+ commits in 2026

See also: AI & agents · Full‑stack & custom software

Offer: AI engineer for your team (Node.js, LLM agents)

Want the number before the call?

$900 to $2,700 a month excl. VAT: 20 to 60 hours at the base rate, $45/h.

See price and start date → Book a free 30‑min call

Good to know

n8n licensing and client work: what the license says

Community edition is for internal use

n8n’s Sustainable Use License allows use only for your own internal business purposes or for non-commercial or personal use (n8n license).

Installing on the client’s server is fine

Setting up and maintaining n8n on a client’s own server is allowed, as long as you do not also sell the hosting (n8n license FAQ).

White-labelling n8n is not allowed

Removing n8n branding and offering it as is is explicitly listed as not allowed (n8n license FAQ).

Clients may not build workflows on your instance

You may build and charge for client workflows on your own instance, but hosting n8n so clients build their own workflows is not allowed (n8n license FAQ).

End users may connect their own accounts

A client’s users may connect their own accounts and credentials to automations you built, as long as they cannot change the workflow logic (n8n license FAQ).

Client chatbots must say they are AI

From 2 August 2026, chatbots must tell people they are talking to AI at the latest at the first interaction (AI Act Article 50).

FAQ

Questions

Will you ever contact our client?
Only if you ask me to. By default I work through you, under NDA.
How do you price agency work?
A monthly package of hours at my base rate, 45 USD, 40 EUR or 180 PLN an hour: 20 hours is 900 USD, 60 hours is 2,700 USD. See pricing. The configurator shows the price for your package.
What if a month needs more hours?
We agree on the extra hours before I start them. Unused hours do not roll over.
Do you work with n8n and Make yourselves?
I read and extend existing workflows, but I am most useful where they end and code begins.
Is there a long commitment?
No. The package is planned month by month, from 20 to 60 hours, and every project ends with a runbook your team can use without me.
Can we start with one client project?
Yes. Hourly at 36 to 54 USD with no minimum, or a one‑week code audit of the client’s app for 1,080 to 2,160 USD, with a go or no‑go you can show the client. The package makes sense once the work is steady.
Who owns the code we deliver to our client?
You do, and you pass it on. Code lives in your repository or your client’s from the first commit, and the contract transfers the copyright to you in writing. AI tools run in the mode you pick, with training off, and client data never reaches a model without your written approval. Before I touch personal data, we sign a DPA.
You are one person. What if you get sick mid-project?
You hear it as early as I can see it, and nothing is stuck with me: code in your repository, decisions written down, a runbook. No subcontractors and no substitution without your agreement, so your client’s code never reaches a stranger. Details on how I work.
Can you cover night incidents for our clients?
No. I do not sell a pager or an SLA, and I say so before you promise one to a client. I reply within 24 hours on working days and keep room for urgent fixes for current clients, mostly in the evenings. What I sell instead is fewer incidents: alerts, backups, deploys you can revert and a runbook your team can follow. See how I work.
Can my agency host n8n for clients?
Not as a service where clients build workflows. Installing it on the client’s own server is fine if you do not also sell the hosting (n8n license FAQ).
Can n8n run inside a product we sell?
As a backend engine that only your own backend talks to, yes, under the Community license (n8n license FAQ). Enterprise features in .ee files need a separate license (n8n license).
Does our client’s self-hosted n8n have a critical vulnerability?
On 8 January 2026 n8n disclosed CVE‑2026‑21858, rated CVSS 10: on versions from 1.65.0 before 1.121.0, certain form workflows let an unauthenticated attacker read files on the server. I upgrade the client’s instance in waves you can revert, put it behind backups and monitoring, and leave a runbook, so your team knows what runs where when something breaks at night. I have run client automations in Make.com and Zapier myself: in 2025 I routed newsletter and contact forms from about ten client sites through a Make webhook with bot checks into one team channel, still running on Make’s free plan a year and a half later. I also put a scheduled trigger behind an authenticated webhook with a lock against parallel runs.

In 30 minutes you will know if we are a fit

You bring the problem, I bring the questions. Want a number first? The configurator shows hours, a price range and the start date.

Book a free 30‑min call Email me instead

or write to ciunkos@gmail.com

After you book you get a calendar invite. Nothing else to fill in.

My pledge

  • Reply within 24 hours. I reply within 24 hours on working days, personally, not a bot. Weekend messages: by Monday noon.
  • Late means free. If I am late, the first hour of work is free. You do not have to ask.
  • Scope in writing within 24 hours. A free 30‑minute call with no company deck, then the scope and the price in writing within 24 hours. You decide.
  • The person who does the work. You talk to the person who does the work. No sales layer, no handoff.
  • Follow‑up only with news. I follow up only when I have something new for you. Never a “did you see my offer”.
  • Honest “not a fit”. If I am not the right person, I say so on the call and point you to someone who is.

Calls this week · work can start now · B2B contract · Wrocław, Poland (CET), remote

You own the code, domain and accounts from day one. Nothing is held hostage.