$450 per workflow, fixed. You write what it has to do; you get an importable JSON file, 72 hours later, with a test harness that shows it surviving retries, duplicate webhooks, missing days and DST. You pay after the tests pass on your data.
Reply time is under a day, usually under an hour. Card or PayPal balance. 14 days to change your mind, no explanation needed.
| Step | What happens | Time |
|---|---|---|
| 1. Spec | You send four things by email: what it must do, which systems it touches, where n8n runs, and the deadline. A paragraph each is enough. No call required. | 5 minutes of your time |
| 2. Quote | I reply with the price, the deadline I can actually hit, and the list of cases I will test. If the job needs two workflows, it says so before you pay. | same day |
| 3. Build | I build and test on a real n8n instance with your sample data (or synthetic data that matches its shape). | 72 hours |
| 4. Handover | JSON, harness, execution output, handover note. You run it; if a test fails on your data, you pay nothing and keep the file. | same day |
Two builds are public with their test output included. Read them first and judge the standard.
| Build | What it proves | Look |
|---|---|---|
| HVAC runtime-hour reminders | One row per customer per day so a retry rewrites the day instead of adding to it; idempotent reset webhook; rebate prompt gated on zip and equipment age. 16 of 16 scenarios pass in a real n8n instance. | gist with execution output |
| Content approval gate | Approvals keyed to a revision, stale links refused as stale_link, idempotent decisions, one 48-hour chase, a publish step that refuses. 20 of 20 scenarios pass. |
live demo · gist |
| CSV repair and dedupe nodes | Encoding, delimiter and duplicate-identity handling that people actually search for. Free, with the failure cases named. | encoding · dedupe |
Free, in the browser, nothing uploaded: paste a workflow JSON into the audit tool and ten mechanical checks run — open webhooks, credentials in node fields, missing retries, writes with no de-duplication, expressions pointing at nodes that were renamed. If it reports nothing, the JSON is structurally sound and the remaining question is the logic, which is what I read.
These come up often enough to answer up front:
Both. The difference is credentials and storage: on Cloud you attach credentials through the UI and can use Data Tables, on self-hosted you point at Postgres. I test on a real instance; if you are on Cloud I will tell you where a Data Table beats a Sheets workaround, and if you are on Postgres you get the schema.
If the API cannot be reached with the HTTP Request node and a Code node, it needs a custom node, which is not a $450 job. I will say so in the quote rather than discover it at hour 60. Most integration work never gets there.
Send a shape, not the data: column names, three synthetic rows, and the edge cases. If the job genuinely needs your rows, I will offer to work under an NDA and delete the copy afterwards, and say plainly what access is needed.
Two weeks of fixes included from handover, for defects in what I delivered. Not for a changed API or a new requirement, which is a new quote, and I say which is which before starting.
PayPal or a card through PayPal, in USD. The charge shows as dragana.karans@gmail.com, which is this business. There is no account to create and no subscription: the checkout is here.
No agency overhead, no account manager, no discovery workshop. You write what it must do, I build and test it. What you lose is hand-holding; what you keep is the client relationship and the margin.
Send the four-line spec. You get a fixed price, a deadline I can hit and the list of cases I will test, the same day. If your job is not a fit, that reply says so too.
Email the spec