The AI and SaaS stack for distributed teams (2026): cost by stage, plus the cross-border catches

Every stack post has the same problem: it lists tools you already know about at prices that were true eighteen months ago. This page does two narrower things instead. It prices a stack by team stage from figures we captured ourselves, and it tells you which of those vendors genuinely behave differently when your company or your card sits outside the United States.

The second part matters because the usual advice is wrong in a specific way. Buying software from outside the US is mostly boring and it works. The friction is real but it clusters in a few places, and it is a payments problem far more often than it is a software problem.

Stage 0 to 1: solo or two founders

At this size the correct stack is almost embarrassingly small, and the main risk is buying process tooling to feel like a company. You need somewhere to write, somewhere to track work, a code host, and an AI coding tool that earns its keep.

Notion's free personal tier is listed at €0, with paid tiers at $10/mo and $8/mo depending on billing period and plan, alongside euro pricing of €9.50 and €19.50 on the localised page. Linear shows a free tier at $0 and paid tiers at $10 per user and $16 per user per month. Cursor lists $20 / mo for individual Pro.

So a two-person team that pays for exactly one AI tool and nothing else lands near $20 to $40 per month. The temptation to add a CRM, a data warehouse and an observability vendor before you have customers should be resisted on cash grounds alone.

Two notes on reading those numbers. Notion's page serves different currencies by region, which is why our capture contains both dollar and euro figures for what is nominally the same plan — if you are comparing quotes with someone in another country you may not be comparing the same number. And Linear's free tier is genuinely usable rather than a trial, which is unusual enough to mention: plenty of teams run on it well past the point where they could afford to pay.

The one place we would spend money early is the AI coding tool, because it is the only item on the list whose output scales with how much you use it. Everything else at this stage is coordination overhead for a team small enough not to need coordinating.

Stage 2 to 5: the first real bill

This is where per-seat pricing starts to compound and where most teams first notice tooling as a budget line rather than a rounding error.

Call it $200 to $300 per month for five people on a disciplined stack. Two observations worth more than the arithmetic.

First, the AI line is the only one that behaves unpredictably. Seats are fixed; model usage is not. A team that leans on agent workflows can exceed its included allowance substantially without changing plan, which is why we suggest setting a per-developer ceiling and reviewing it monthly rather than annually.

Second, this is the stage where a payment method becomes load-bearing. Five subscriptions renewing monthly on a founder's personal card is a single point of failure, and the failure mode is a declined renewal on a Friday rather than anything dramatic.

Stage 5 to 15: where the shape changes

Past roughly ten people the stack stops being a list of tools and starts being a set of decisions about who administers them. The licence cost grows linearly and predictably. What grows non-linearly is access management, offboarding, and the number of vendors holding your payment credentials.

Three costs appear at this stage that nobody budgets for. The first is seats for people who no longer need them, which is the single most common source of waste in a growing team's tooling bill — every tool bills per seat and nobody enjoys auditing them. The second is the higher tier you get pushed into for one feature, usually SSO or an admin control, which can move a $10 seat to $16 across the whole team to solve a problem for one person. The third is duplicated function: a docs tool, a wiki inside the issue tracker, and a shared drive all holding overlapping content.

None of that is a cross-border problem, and we are flagging it anyway because it is the part of the bill you can actually control. The vendor-selection question at fifteen people is much less interesting than the administration question.

At this size the same three tools cost somewhere between $600 and $900 per month at list, and that is usually still not the biggest line in the budget — payroll and contractor costs dominate, which is the subject of our EOR pillar. The reason to care about tooling here is not the absolute number but the fact that nobody owns it. Assign one person.

The cross-border part, honestly

Here is the claim we want to make carefully, because the internet is full of confident nonsense in both directions.

Buying mainstream SaaS from outside the US mostly just works. Self-serve seat subscriptions at Linear, Notion and Cursor bill to non-US companies and non-US cards. There is no hidden gate that requires a Delaware entity to pay $10 a month for an issue tracker, and anyone telling you otherwise is selling incorporation services.

The friction that does exist concentrates in three places:

1. The card, not the country. The common failure is a card that a vendor's processor declines or that your own bank blocks for recurring foreign-currency charges. This is why the fintech layer sits underneath the tooling layer: solve it once with an account that issues cards vendors accept and you stop solving it per tool. Airwallex lists unlimited multi-currency corporate cards with zero international fees on its $0-per-user Explore tier and up to 2% cash rebates on local USD spend, and Mercury lists a $0 monthly tier — both captured 2026-08-03. Detail in the fintech pillar.

2. The billing entity, for anything above self-serve. Enterprise agreements, invoicing on terms rather than card, and some payment-processing and app-store relationships are where a US entity genuinely changes what is available to you. That is an argument for incorporating when you hit those needs, not before. See the incorporation pillar.

3. Partner credit programmes, which are worth real money. Stripe Atlas states a $500 one-time fee including government fees and the first year of registered agent service, and bundles $2,500 in Stripe credits plus over $50,000 in stated partner discounts on tools including Mercury, Xero and AWS. Whether that discount bundle is worth anything depends entirely on whether you would have used those vendors anyway — most of it is credit against spend, not cash.

What we deliberately do not claim: that specific tools are unavailable in specific countries. Vendor availability lists change, sanctions regimes change, and a page that asserts "tool X is blocked in country Y" ages badly and dangerously. When it matters for your jurisdiction, check the vendor's own terms on the day you buy.

Per-vendor notes for teams billing from outside the US

The table below separates two things that get conflated: what the vendor's own pricing page states, and what we have actually verified ourselves. We are being pedantic about the difference on purpose, because most cross-border restrictions content is confident about things nobody checked.

Two things are absent from that table deliberately. We have not verified per-country KYC outcomes for any of these vendors, because that depends on your jurisdiction, your documents and their risk appetite on the day — a page cannot tell you that truthfully. And we have not listed AI model vendors' regional access rules, which is a large enough question that it gets its own page.

The comparisons people actually search for

Four questions come up repeatedly, and each deserves more room than a pillar page can give it.

Cursor versus GitHub Copilot. This is a usage-model question rather than a seat-price one. We have a captured price for Cursor and not for Copilot, so we are not going to print a head-to-head number here; doing so from memory is exactly how stale pricing spreads. The structural point holds regardless: compare what each includes before overage, not the headline seat.

Linear versus Jira for a small team. For teams under about fifteen people the deciding factor is rarely features, it is how much configuration the tool demands before it is useful. Linear's free tier lets you answer that question at zero cost.

Notion versus Linear. These are not really competitors. Teams that treat them as one end up with issues in a wiki or documents in an issue tracker. If budget forces one, pick based on whether your bottleneck is writing things down or shipping things.

Stripe alternatives for a non-US business. This is the highest-intent question in the cluster and it is a payments question, not a tooling one. It belongs in the fintech pillar, where we cover collection accounts and FX spreads with captured pricing.

Commercial versus open source

We run our own stack page and we are not neutral: we pay for tools where the alternative costs us hours. The test we apply is simple. If nobody on the team is paid to operate infrastructure, self-hosting is not cheaper — it moves a $10 licence into an unbudgeted maintenance obligation. Once you have someone whose job includes running things, specific swaps start to make sense.

This is where our sibling site swaptosaas.com goes deeper on open-source replacements. The division of labour: they cover cost-cutting by substitution, we cover the commercial stack with cross-border operating notes.

A budget you can defend

If you want one number to take into a planning conversation, use this: $40 to $60 per person per month covers issue tracking, documentation and an AI coding tool at list prices we captured on 2026-08-03. Everything beyond that is either infrastructure, which scales with usage rather than headcount, or a tool someone asked for and nobody reviewed.

The reason to write the number down is that tooling spend fails quietly. It does not arrive as one bad decision; it accumulates as fifteen reasonable ones, each too small to argue about. A per-head figure gives you something to compare against, and a monthly review gives you the moment to do it.

Three habits do most of the work. Review seats monthly rather than annually, because seat drift is the largest recoverable waste in any growing team's bill. Keep the AI usage line separate from the seat line in whatever you track spend in, since they behave differently and averaging them hides the volatile one. And put subscriptions on a company card with a real limit rather than a founder's personal card — not for accounting neatness, but because the alternative fails at the worst possible moment.

What we would change about this page

Two gaps we know about, stated plainly rather than papered over. We do not have captured pricing for GitHub Copilot, Slack or Google Workspace, all of which belong in an honest stack page — they are not in our snapshot pipeline yet, so they are not in our tables. And our cost-by-stage figures assume list price, which understates what a fifteen-person team pays once one tool pushes it onto a higher tier for a single feature.

We would rather ship a page with visible gaps than fill them from memory. When those vendors enter the pipeline, this page gets updated with the capture date changed, and the numbers above will be the ones we actually fetched.

What to do next

If you are under five people: buy one AI tool, use free tiers for everything else, and revisit in a quarter. If you are between five and fifteen: name an owner for tooling spend, get the subscriptions off a personal card, and set a monthly ceiling on AI usage. If your problem is that a vendor will not take your money, that is a fintech problem — start at the fintech pillar, not here.

Prices captured from vendor pricing pages on 2026-08-03 and stored as snapshots. Some links on this page are affiliate links, marked rel="sponsored"; they do not change what we recommend or the prices you pay.