# The real monthly cost of Clerk plus Stripe plus Lago plus Knock

> Four best-of-breed vendors, added up with the assumptions shown. The subscriptions are not the expensive part.

- Source: https://www.buildbase.app/blog/auth-billing-stack-cost
- Section: Blog
- Author: Dharmendra Jagodana
- Published: 2026-08-14

## In short

At 20,000 users the published subscriptions for auth and notifications come to roughly $375 a month before Stripe fees, and metering is unpriced because Lago publishes no cloud rate. The larger cost is the integration you own between four systems, which no pricing page shows and which does not go away.

Every "build vs buy" post compares subscription prices and stops there. The
subscriptions are the part you can see. They are rarely the part that makes the
decision expensive.

Here is the arithmetic anyway, with every assumption stated so you can argue
with it.

## The assumptions

Change any of these and the total moves. That is why they are listed.

- A B2B SaaS with **20,000 monthly active users**
- Selling a **metered** product, so usage has to be counted and billed
- **Transactional notifications**: receipts, alerts, quota warnings
- US-based, standard card payments
- No more than 100 customer organizations, so Clerk's per-organization
  metering ($1 a month for each one past 100 per app) never applies
- Prices verified **24 September 2026** and quoted from each vendor's own page

## The subscriptions

| Vendor | Job           | Monthly at 20,000 users                |
| ------ | ------------- | -------------------------------------- |
| Clerk  | Auth          | $25 Pro plus $100 B2B add-on, so $125  |
| Stripe | Payments      | Percentage of revenue, see below       |
| Lago   | Metering      | Not published, see below               |
| Knock  | Notifications | $250 Starter, 50,000 messages included |

**Clerk** is the pleasant surprise. The Pro plan is $25 a month with 50,000
monthly retained users included, and $0.02 for each one beyond that. At 20,000
users you are inside the included allowance. The costs that bite are the add-ons:
each enterprise SSO connection beyond the first is $75 a month, and the B2B
organizations add-on is $100 a month. A B2B product selling to teams probably
needs the second one, which takes the auth line from $25 to $125.

**Knock** is the largest published line at $250 a month for Starter, including
50,000 messages, then $0.005 per message. A metered product that emails a quota
warning at 80 percent, a receipt monthly and an alert on failure will generate
more messages than user count suggests.

**Stripe** is a percentage rather than a subscription. It scales with revenue,
not with users, which makes it the one line you should be happy to see grow -
though if it's specifically the Billing add-on's platform fee pushing you to
look elsewhere, [Stripe Billing alternatives](/compare/stripe-billing-alternatives)
sorts the options by what they actually replace. And if it's the revenue-share
shape itself you're weighing rather than Stripe specifically, [the crossover
point where a percentage fee costs more than a flat
plan](/blog/billing-platform-revenue-tax) runs that math with your own
numbers.

**Lago** publishes no cloud pricing. The self-hosted edition is open source
under AGPLv3, so the real cost there is infrastructure and the engineer who maintains
it. Cloud pricing requires a sales conversation. I am not going to invent a
figure for a post whose entire purpose is arithmetic you can check. If metering
depth specifically is what's pulling you toward Lago rather than the rest of
this stack, [how it stacks up against bundling metering into
BuildBase](/compare/lago-vs-buildbase) is the narrower version of this same
arithmetic.

So: **roughly $375 a month** in visible subscriptions at this shape, or $275 if
you skip the B2B add-on, plus Stripe's percentage, plus whatever Lago costs you.

## The line nobody puts on a pricing page

Four vendors means three integrations you own forever.

Clerk knows who the user is. Lago knows what they used. Stripe knows what they
paid. Knock knows what they were told. Nothing in that list knows all four
things, so you write the code that makes them agree, and you keep writing it
every time one of them changes an API or a webhook payload.

Concretely, the agreements you maintain:

- A user in Clerk maps to a customer in Stripe maps to a subscriber in Lago
- Usage recorded in Lago has to reach Stripe before the invoice closes
- A quota warning needs usage from Lago, entitlement from Lago or Stripe, and
  identity from Clerk before Knock can send it
- Every one of those has a failure mode, and the failure mode is a wrong invoice

That is not hypothetical overhead. It is the code that produces the support
ticket that starts "I was charged for".

## Where the assembled stack wins

Honestly: in depth, and in exit.

Each of these products is better at its single job than a module in a bundle.
Knock's notification engine does more than our email module. Lago's metering
model is richer. If one of them stops fitting, you replace one thing rather than
migrating a platform.

If your product's difficulty is concentrated in exactly one of those areas, buy
the best one for that area and assemble around it. That is a good decision and
this post is not arguing against it.

## Where the bundle wins

When no single area is your hard problem, and you need all four to work together.
Then the integration is the work, and it is work that produces no product.

Our numbers are at [/pricing](/pricing). The comparison against a vendor that
takes the assembled-stack position is in
[Kinde vs BuildBase](/compare/buildbase-vs-kinde), the entitlements-first
position in [Stigg vs BuildBase](/compare/buildbase-vs-stigg), and the
identity-first position in [Auth0 vs BuildBase](/compare/auth0-vs-buildbase),
and the auth-with-billing-attached position - Clerk is the first line in the
table above - in [Clerk vs BuildBase](/compare/clerk-vs-buildbase). There is
also a version of this argument one layer lower, starting from a raw database
instead of a vendor stack, in
[Supabase vs BuildBase](/compare/buildbase-vs-supabase), and the
mobile-and-realtime-first version of that same primitives argument in
[Firebase vs BuildBase](/compare/firebase-vs-buildbase). There is also a
no-code, page-builder-first version of the same choice, for a stack built on
Webflow or Squarespace instead of a codebase, in
[Outseta vs BuildBase](/compare/outseta-vs-buildbase). And the same
sticker-price-versus-real-cost pattern shows up in pure infrastructure terms,
no vendor stack involved, in
[what database-per-tenant actually costs](/blog/database-per-tenant-cost), and
again in the self-hosted-versus-managed decision itself, where the ops hours
replace the subscription line entirely, in [Self-host vs managed: the real
cost comparison](/blog/self-host-vs-managed-cost).

If your plan was to skip the vendors entirely and buy a boilerplate, the same
four accounts show up there too -
[what a one-time boilerplate leaves you operating](/compare/shipfast-vs-buildbase)
is where that version of the arithmetic goes. Skip the vendors and build the
pieces yourself instead, and
[the auth piece alone still costs real hours](/blog/cost-of-building-auth-yourself)

- [metering runs a similar
  bill](/blog/cost-of-building-usage-metering-yourself), most of it recurring as
  maintenance rather than the initial build. Not four subscriptions this time -
  your own engineering time.

## Re-check these numbers

Every figure above has a date on it because every figure above will move. Clerk,
Knock and Stripe all changed pricing in the last two years. If you are making a
decision on this post more than a month after that date, open the four pricing
pages yourself. That is the honest instruction, and it applies to our page too.
