Choose notification infrastructure by who controls the routing decision, what users can suppress, and what the subscription counts. Novu suits a workflow-centered implementation with an open-source deployment path; Knock offers broad cross-channel workflows and built-in multi-tenancy; Courier merits a shortlist for usage-based sending without a fixed Business subscription; SuprSend makes advanced preferences and data modeling explicit package decisions; MagicBell offers a focused inbox-led starting point. These are conditional fits, not a delivery-speed ranking.
The buying problem is an application that must coordinate an inbox, email and other channels after a business event. An email transport alone does not decide whether an already-active user should also receive a reminder. Our transactional email comparison addresses that downstream transport purchase; our SMS API comparison addresses carrier-facing sending. Keep those provider costs alongside the orchestration bill.
Official sources checked October 24, 2026. Dollar amounts below are displayed USD platform prices. The same event can create multiple recipients, workflows and channel deliveries, so the allowances are not interchangeable.
Compare the unit before the allowance
| Product | Useful fit | Published commercial reference | Decisive qualification |
|---|---|---|---|
| Novu | Workflow-centered routing and deployment choice | Pro from $30/month; 30,000+ workflow runs | One subscriber execution, across environments; provider charges separate |
| Knock | Cross-channel workflows with tenant context | Starter $250/month; 50,000 messages, then $0.005 each | Guides and agent credits have separate allowances |
| Courier | Usage-based sending without Business base fee | Developer 10,000 sends free; Business $0.005/send | Confirm send accounting for the actual multichannel route |
| SuprSend | Preferences, digests and recipient data modeling | Business $250/month billed annually at $3,000 | 50,000 notifications; additional $5/1,000; downstream providers extra |
| MagicBell | Inbox-led delivery with direct channel integrations | Startup $249/month; 50,000 deliveries | $0.0025 per additional delivery; channel fan-out counts |
Novu
Novu’s current plans start with 10,000 monthly workflow runs, up to 20 workflows, two environments and three team members on Free. Pro starts at $30/month with 30,000+ runs and seven-day activity retention. Team starts at $250/month with 250,000+ runs, more environments, longer activity retention and RBAC. A low starting amount should not conceal the governance package a larger team needs.
The billing definition is particularly useful: a run executes one workflow for one subscriber, regardless of its steps or channels. Development and production usage aggregate at organization level. Topic fan-out creates a run for each matching subscriber. Several channel messages can therefore belong to one billed run; sending to a large audience is not one run. Cloud charges also exclude email and SMS provider consumption.
Prefer Novu when the workflow is the unit your engineering team wants to own and reason about. Its no-license-cost open-source self-hosting path differs from enterprise self-hosting with commercial support and business features. Owning deployment adds an operational responsibility; a free software core is not the same purchase as managed Cloud.
Knock
Knock combines reusable workflows, channel routing and recipient data in its platform model. Its pricing matrix includes unlimited workflows, stored profiles, team members and built-in multi-tenancy in Free. Starter costs $250/month with 50,000 messages and a $0.005 additional-message rate. It defines a message as delivery to one user on one channel and excludes messages that bounce or fail to send.
The important budget boundary is that messaging is not the entire product bill. Guides count unique recipients seeing an in-app banner, popup or announcement, while agent credits have their own allowance and rate. A company adding a launch announcement to an existing notification implementation should estimate those meters separately. Built-in tenancy also does not imply every per-customer branding or permission feature is included in the entry package; those appear in the Enterprise offer.
Knock is a strong candidate when shared workflows need tenant context and the team wants to extend into other in-product communication. If all you need is a small unread inbox, that broader scope may add little value.
Courier
Courier’s current offer lists Developer with 10,000 sends/month and Business at $0.005 per send, without per-seat or channel add-on fees. Enterprise adds configured volume pricing, access controls, observability integrations and an SLA. The lack of a Business base subscription makes it worth evaluating for a workload that does not justify a fixed monthly platform fee.
Its Inbox documentation supplies a separate in-app delivery surface. That does not establish that a workflow run, provider attempt and billed send are equivalent. Before budgeting fan-out or fallback routes, obtain the applicable definition of a send and the treatment of development traffic and failures. The retrieved pricing FAQ exposed those questions without sufficient answer text, so an exact multichannel invoice is not inferred here.
Choose the usage model only after checking that the required preferences, retention, region and operator roles are in the purchased package. A low-volume application with strict access requirements may still need Enterprise.
SuprSend
SuprSend’s annual pricing view distinguishes Essentials at $100/month, billed $1,200 annually, from Business at $250/month, billed $3,000 annually. Both include 50,000 notifications, but Business adds batching, digests, user preferences and object data modeling. Their additional-notification rates differ: $2 versus $5 per 1,000. The same included volume does not buy the same workflow.
The preferences model supports category and channel choices. This matters when a user may suppress a collaboration update while retaining another category. The price FAQ counts channel notifications separately, excludes messages suppressed by SuprSend, and counts a successfully triggered notification even when the downstream provider fails. Email and SMS provider accounts remain a separate expense.
SuprSend fits a team that needs these recipient controls as part of its communication model. Confirm multi-tenant entitlements separately: the matrix also marks tenant customizations as an add-on. A trial exposing premium features is not evidence that the chosen production tier includes them.
MagicBell
MagicBell’s documentation starts with selecting a user identifier, adding the inbox and broadcasting. It also documents email, push, SMS, Slack and other channel integrations. The useful shortlist reason is a focused inbox integration, rather than an assumption that the service only supports in-app messages.
Its Builder plan includes 1,000 deliveries/month and one project. Startup is $249/month for 50,000 deliveries and five projects, with $0.0025 additional deliveries and $99 additional projects. Deliveries add up across the channels a user receives, subject to preference or smart-delivery suppression. Unlimited users therefore does not mean unlimited delivery volume. Confirm downstream channel costs and the required security/support offer before using the platform price as a total.
Make suppression and ownership explicit
One business event is not one billable notification. Shortlist against the actual route: eligible recipients, tenant permissions, channel preferences, batching, fallback, log retention and who can change a template. Suppression can be correct behavior, rather than a delivery defect. Conversely, provider acceptance does not prove that a customer read a message. Keep the application event authoritative and define what support staff need to inspect when a user reports a missing notification.
A single-channel application may be better served by its existing transport and a small application-owned preference model. Add orchestration when coordinated routes, user control and operational visibility justify a second system.
By Toolverly Editorial, with AI-assisted research and drafting.
Photo by Rob Wingate on Unsplash. Read about our publication or send a correction.