BTCPay Server vs NOWPayments

Same custody model, opposite operating burden#

Both settle to a wallet the merchant controls, which is the property most merchants say they want and few actually get. Neither can freeze a balance, because neither carries one on your behalf.

After that they diverge completely. BTCPay Server is a program you install on a server you rent, patch and monitor. NOWPayments is a hosted endpoint you call, with the uptime problem living on somebody else's rota.

That distinction is the entire comparison. Everything below is a restatement of it in some particular currency: money, assets, integration effort or attention.

Zero against one percent, and what fills the gap#

The software charges nothing at all. The service charges a flat 1% on processing, published and applied identically from the first payment onward.

One percent of a hundred thousand a year is a thousand. A modest virtual server plus a few hours a month of somebody competent is not obviously cheaper than that, and it is considerably less predictable.

The arithmetic flips at volume. At a million a year the fee is ten thousand and the server still costs what a server costs, which is when running your own stops being ideology and starts being finance.

Two networks against three hundred and fifty assets#

The self-hosted option documents Bitcoin on-chain and Lightning and publishes no asset count at all; altcoin support exists through separate plugins whose status moves independently of the core release.

The hosted option publishes the longest asset list in this index. That is not a marginal advantage for a merchant whose customers arrive holding whatever they happen to hold, and it is worth nothing to a business paid in bitcoin by people who wanted to pay in bitcoin.

Decide which of those two sentences describes your customers before weighing anything else here. The answer settles the pair more often than the fee does.

Plugins are the surprise on this pair#

The free software names twenty-three ecommerce integrations on its own page. The paid service enumerates none, leaving the API as the documented path.

That inverts the usual expectation, and the reason is structural rather than accidental: a community writes the plugin for the platform it personally runs, while a company writes the plugins its sales pipeline justifies.

For a shop owner on WooCommerce or Shopify that gap is the whole project plan. For a team with engineers it is noise, since they were going to call the API in either case.

Neither one gives you euros#

Fiat settlement is documented as absent on one card and recorded as unpublished on the other. In practice a merchant on either side receives crypto and arranges conversion through a second relationship.

This follows from the shared custody model rather than from any decision either provider could reverse. A system that never holds your balance is not the shape that pushes bank transfers on a schedule.

If bank currency on a timetable is a requirement, this pair is the wrong page and the fiat-settling providers are where to start instead.

Onboarding is nothing against unknown#

There is no verification step on the self-hosted side because there is no counterparty to perform one. You download software; nobody approves you.

The hosted service publishes no merchant verification requirements whatsoever. A non-custodial design structurally has less to check, which is the reasoning behind its onboarding score, but an undocumented process should be planned for as slower rather than lighter.

Neither publishes a jurisdiction. One is an open-source project with no company at all, and the other is a company that declines to say where it is.

What to ask each of them#

To the project: which altcoin plugins are actively maintained against the release you intend to run, and who maintains the one you need. The list is long and its maintenance is uneven, and that unevenness is invisible from the feature page.

To the service: the conversion spread and the payout cost split into provider fee and network fee. Two of its three cost components are unpublished, and the flat 1% is only the part it chose to show.

To yourself: who patches the server in eighteen months, and what happens the week that person is on holiday and an invoice stops confirming.

Where this pair actually lands#

A Bitcoin-denominated business with an engineer takes the software, pays nothing, and accepts an operational commitment measured in years rather than weeks.

A merchant meeting unpredictable customers, or one without anybody to own a server, takes the service, pays 1%, and keeps the property that matters most about both cards.

The uncomfortable middle is a small shop that wants zero fees and has nobody to run anything. That business should pay the one percent and stop reading, because the cheapest option here is only cheap for somebody who can afford the time.

The failure modes are not the same size#

A hosted service failing takes your integration offline until it comes back, and the incident is somebody else’s to resolve while you answer customers. That is unpleasant and it is bounded.

A self-hosted node failing is unbounded in a different direction: nobody is paged unless you built the paging, and a Lightning channel that cannot route at four in the afternoon on a busy day is a problem with your name on it.

Neither is worse in the abstract. What decides it is whether your business would rather buy a service level it cannot verify or own a problem it can actually fix.