Knock Pricing in 2026: What Notification Infra Really Costs

Notification pricing pages quote a sticker. The sticker is not the bill. On notification infrastructure, the bill is set by the meter: the unit each vendor counts. Pick the wrong meter for your channel mix and you pay three times over for the exact same alerts.

We priced three notification platforms head to head as of July 2026: Knock, Courier, and Novu. Same workload, three different meters, a 14x spread on the monthly bill.

What does Knock actually cost in 2026?

Here is Knock's public pricing as of July 2026, from its pricing page:

The headline number is clean: $250 per month for 50,000 messages, then half a cent each. The problem is the word message.

The hidden meter: Knock bills per channel, not per notification

Knock defines a message as "a single message delivered to a single user on a single channel." Read that again with your own product in mind.

Say you send an order-shipped alert. It goes to email, to mobile push, and to the in-app feed. To you that is one notification. To Knock's meter it is three messages. Your bill scales with this formula:

> billable messages = users x notifications per user x channels per notification

The channel count is a multiplier most teams forget when they read the sticker. A per-notification mental model meets a per-channel-delivery meter, and the invoice arrives at triple what the pricing page implied.

A real 30-day bill: 5,000 users, 3 channels

Let us put a concrete workload through it. A mid-size SaaS with:

• 5,000 monthly active users
• 10 notifications per user per month
• 3 channels per notification: email, push, in-app feed

That is 50,000 notification events, and 150,000 channel deliveries.

On Knock, the meter counts channel deliveries, so the billable unit is 150,000 messages:

• Starter base: $250 for the first 50,000
• Overage: 100,000 x $0.005 = $500
• Knock total: $750 / month

None of that $750 is exotic. It is 150,000 half-cent deliveries with a base fee on top. The sticker said $250. The channel multiplier said $750.

Knock vs Courier vs Novu: same workload, three meters

Now run the identical 50,000 notifications through two competitors that count differently.

Courier bills per send and states plainly that "pricing is based on notifications sent, not team size or channels." Read at face value, a multi-channel notification is one send. So the billable unit is 50,000 sends, not 150,000:

• Business pay-as-you-go: 50,000 x $0.005 = $250 / month

Novu bills per workflow run. A run is one execution for one subscriber, and the channel steps live inside that single run. So the billable unit is again 50,000, not 150,000:

• Pro base: $30 for 30,000 runs
• Overage: 20,000 runs at $1.20 / 1,000 = $24
• Novu Pro total: $54 / month

Same alerts. Same users. Same three channels. Here is the scoreboard:

That is a 14x spread, $54 to $750, driven almost entirely by which unit the vendor decided to count. One honest caveat before you screenshot this: confirm each vendor's exact multi-channel definition against your own routing, because a "send" or a "run" that quietly splits per channel would change these numbers. The pricing pages above are the source of truth as of July 2026.

The one lever: match the meter to your fan-out

The cheapest platform is not a fixed fact. It depends on your channel count.

• If you fan out to many channels, a per-notification meter (Courier) or a per-run meter (Novu) is dramatically cheaper, because neither multiplies by channels.
• If you send single-channel alerts, the meters converge and Knock's polish may win outright.

So the lever is not "pick the cheap vendor." It is "stop paying the channel multiplier where you do not need it." If 40 percent of your volume is a low-value in-app feed that duplicates the email, drop it or route it off the per-channel meter. On our workload, going from three channels to two takes Knock from 150,000 to 100,000 messages:

• $250 base + 50,000 x $0.005 = $500 / month

One config change, 33 percent off, no vendor migration required. The realtime in-app layer specifically is often the channel worth pulling in-house; we ran the same exercise on Pusher's connection meter and the pattern repeats.

When Knock is worth $750 a month

This is not a cheapest-wins verdict. Knock's bill buys real things Novu self-host makes you build and operate: a managed in-app inbox, a hosted preference center, template management, delivery tracking, and a dashboard your non-engineers can use. Courier sits in the middle with a strong routing and template layer.

If notifications are a core product surface and engineering time is your scarce resource, paying per channel can be cheaper than staffing the equivalent internally. The honest failure mode is the reverse: a high-volume, low-value notification stream (activity feeds, digests, marketing nudges) sitting on a per-channel-delivery meter, quietly compounding.

Build vs buy: self-hosting Novu

Novu is open source. You can self-host it from its GitHub repo and pay only for the server plus your own operations time. At 50,000 runs that is a rounding error against $54 of managed Pro, but at 5 million runs the managed bill is real money and self-host starts to pay for itself.

The trade is the usual one, and it is the same trade we walked in build vs buy at scale: managed platforms sell you time, not just software. Self-hosting removes the per-run bill and hands you uptime, the inbox UI, and the preference center as your problem. Buy when notifications are adjacent to your core and you want the vendor to absorb the upkeep. Build when notification volume is core, high, and cost-sensitive enough to fund a maintainer.

Developers have flagged this cost cliff for years. As one Hacker News thread from November 2022 put it, "lot of products like Knock, novu, engagespot, courier, the pricing is on the upper side." The tools have improved since; the meter math has not changed.

Math check: on a per-channel-delivery meter your bill is users x notifications x channels. On a per-notification meter it drops the x channels. That one missing multiplier is the difference between $750 and $250 for the exact same alerts.