Per-seat pricing punishes small teams. Here is why we price by chats instead.

Every support tool I tried before building Svellyo had the same shape of bill. A price per seat, per month. Sometimes a second price per website. Often a third line for the AI features, which were always in the tier above the one I was on.

For a support department that is a fine model. The number of agents is the thing that scales, and every agent costs the vendor real infrastructure and real onboarding. For a founder running two or three small products with a co-founder and a part-time helper, it is a punishment for exactly the behaviour you want: install the widget everywhere, invite everyone who might answer a question, and let the software absorb the volume.

This post explains the model we chose instead, exactly how a chat is counted, and the trade-offs that came with it.

The problem with seats when you are three people

Suppose you and a co-founder run three products. Each product has a marketing site and an app. With a typical per-seat, per-site tool you are looking at two seats times three workspaces, and the AI add-on if you want it to answer anything. That is six paid seats for two people, before you have handled a single conversation.

The natural reaction is to cut corners. You install the widget on only the biggest product. You do not invite your co-founder, so all the chats land on you. You skip the AI tier and answer "how do I reset my password" by hand for the four hundredth time. The tool has quietly trained you to use it less.

Seat pricing also has nothing to do with the value you get. Ten chats a month or a thousand, the bill is the same. The tool's cost to serve you is almost entirely a function of conversations, not of how many people can log in. Charging by seats means light users subsidise heavy ones, and small teams subsidise everyone.

What we charge for instead

Svellyo has one meter: active chats per month, counted across every workspace you own.

• Free: 50 chats a month.
• Pro: 200 chats a month, for $29 a month or $290 a year.
• Scale: 1,000 chats a month, for $79 a month or $790 a year.

Everything else is either included or a plan feature rather than a unit of billing. Workspaces are unlimited on every plan, including Free. Teammates are included, with a per-workspace seat limit that is generous on Pro and unlimited on Scale. The AI agent is included on every plan. The help center is included on every plan, with a custom domain on Pro and Scale.

If you run three products, you create three workspaces. Your chats from all three add up against one limit, and that is the entire bill.

What exactly is a chat

We spent more time on this definition than on anything else in billing, because a vague meter is worse than a seat.

A conversation counts once per calendar month, in UTC, if a visitor engaged with it that month. Engaged means the visitor sent at least one message, or tapped "talk to a person". Some consequences of that rule, written out so there are no surprises:

• A visitor who sends fifteen messages in one conversation in March is one chat.
• If the same conversation continues in April with new visitor messages, it is one chat in March and one in April. It is not fifteen chats, and it is not zero.
• A conversation where only your team wrote in a given month, for example a follow-up you sent, counts for nothing that month.
• Opening the widget, browsing the help center tab, or opening the composer and closing it again does not create a conversation and does not count.
• Messages left through the offline "leave a message" form count, because a person wrote to you.
• AI replies never count. Neither do your replies, internal notes, or AI drafts.
• Deleting a conversation from your inbox does not reduce the count. Usage reflects what your visitors did, not what you kept.

That last one deserves a sentence. Early on we had usage tied to the conversation row, so deleting a thread quietly lowered the meter. It felt generous until we realised it made the number meaningless. Now the meter is its own record. Delete whatever you like from the inbox and the month's count stays honest.

Why chats and not messages, or AI tokens

Messages are the wrong unit because they punish good conversations. A thorough back-and-forth that solves a problem should not cost ten times a curt one.

AI tokens are the wrong unit because nobody can predict them and nobody wants to. You should be able to look at your inbox, count the people who wrote to you this month, and know the bill. We do show AI spend in Analytics, and there is a daily cap as an abuse guard, but it is information, not a line item.

Conversations map to the thing you actually care about: how many people needed help. If that number grows, your business is growing and the bill should grow gently with it. If it does not, you should not be paying for seats that sit idle.

What happens at the limit

Nothing dramatic, and nothing is deleted.

When you reach your monthly limit, the AI stops replying to new visitor messages and the widget switches to a "leave a message" form for new conversations, so you still capture the lead. Your team can keep replying to every conversation from the inbox, because you should never be locked out of talking to your own customers. On the first day of the next month, or the moment you upgrade, everything resumes.

You get an email at 80 percent and at 100 percent. The usage meter is on the home page of every workspace and broken down per workspace on the billing page.

Trade-offs we accepted

A meter this simple has costs, and it is fair to name them.

Big teams pay less than they might elsewhere. A ten-person support team on Scale pays $79 a month. We are fine with that. Those teams are not who we built this for, and the ones that show up will tell their friends.

Usage is aggregated per owner, not per workspace. If you own three workspaces, they share a limit. If a teammate owns a workspace you merely belong to, that workspace counts against their plan. This is simple to reason about and matches how indie makers actually organise things, but it means that when a product changes hands you should transfer the workspace, which we support in one click.

No overage billing. We could have charged per extra chat past the limit. We chose the soft limit instead because a surprise invoice after a launch day is the worst possible experience, and a limited widget that still collects messages is a better one. If you hit the limit regularly, the upgrade button is right there.

Free is genuinely free, and it is limited. Fifty chats a month is plenty for a side project and not enough for a product with real traction. That is on purpose. We would rather you outgrow Free than resent it.

Where the money goes

Since we are being specific: each AI reply costs us a fraction of a cent in model fees, and each conversation costs storage, realtime delivery and a little compute. At 200 chats a month, Pro covers its costs with room to spare. At 1,000 chats, Scale does the same. We are not subsidising usage with venture money, which is why the prices are unlikely to move against you.

The short version

You pay for the number of people who wrote to you. You do not pay for products, for teammates, for the AI, or for a help center. Deleting things does not lower the bill and hitting the limit does not lock you out. If you have ever done the seat maths and decided to skip live chat on your second product, this is the model that makes that decision go away.

The full comparison is on the pricing page, and the counting rules are written up in the help center so you never have to take a blog post's word for it.