DEV Community: Alexandre Paula

7 Firebase Cloud Messaging Alternatives in 2026 (Honest Guide)

Alexandre Paula — Tue, 03 Nov 2026 14:30:00 +0000

Originally published on the QueueFly blog — canonical URL points to the original.

Search "Firebase Cloud Messaging alternatives" and you'll get a wall of vendor listicles that all quietly dodge the hard part: FCM isn't one thing, so "alternatives" isn't one question. Before you shortlist anything, decide which of these you're actually asking:

"I want to stop depending on Google." — a transport question (APNs, self-hosted, UnifiedPush). Only web push and Android-without-GMS apps can really answer this, and the honest options are short.
"I want more than a transport." — a stack question: campaigns, segmentation, analytics. That's the marketing-platform lane (OneSignal and friends).
"FCM works, my delivery code around it doesn't." — a queue layer question: retries, token cleanup, scheduling, fan-out. This is the question most listicles never name, because almost none of them are built to answer it.

I've built on all three lanes — 15+ years of backend work, and lately my own product, QueueFly, sits squarely in lane three. Here are the seven real options in 2026, sorted by which question they answer. Pricing was checked against official pages in October 2026.

Quick picks

Best if you want to keep FCM and delete your plumbing code: QueueFly
Best all-in-one replacement for the whole notification stack: OneSignal
Best for teams already on Pusher: Pusher Beams
Best multi-channel orchestration: Courier
Best for notification-heavy product apps: Knock
Best genuinely self-hosted option: ntfy
Best for cloud-scale plumbing you want to own: AWS SNS

The comparison table

Option	Lane	You still own	Pricing shape (verified Oct 2026)
QueueFly	3 — queue layer over FCM	Nothing beyond your Firebase project	Flat €18–€134/mo by device count, unlimited sends on every plan; 30-day trial
OneSignal	2 — full stack	Nothing; it replaces your transport config	Free ≤1,000 mobile MAU; Growth from $19/mo + ~$0.012/MAU
Pusher Beams	3 — managed FCM/APNs	Deeper queue/DLQ machinery	Free 1k subscribers; tiers $29–$399/mo
Courier	2 — orchestration	Channel strategy	Free 10k sends/mo; then $0.005 per send
Knock	2 — product notifications	Transport credentials	Free 10k msgs; Starter $250/mo
ntfy	1 — transport replacement	Your servers, scaling, mobile UX	Free (OSS)
AWS SNS	1 — transport	Queues, retries, tokens, scheduling, everything	~$0.50 per million push deliveries after 1M free monthly requests

First, the awkward truth about "replacing FCM" on Android

On iOS you have exactly one transport: APNs. Every "FCM alternative" for iPhone apps is still delivering through Apple's gateway.

On Android, FCM is the delivery path for devices with Google Play services — which is effectively all consumer devices. Alternatives like MQTT-over-your-own-server or UnifiedPush exist and matter for de-Googled setups (Huawei HMS, /e/OS, F-Droid-first audiences), but for a mainstream app, "we replaced FCM on Android" usually means "we replaced it with something less reliable." The vendors selling lane-1 and lane-2 options know this — which is why most of them quietly use FCM underneath. That's the dirty little secret of this whole keyword: most FCM alternatives are FCM with a UI.

So the practical question isn't "FCM or not." It's "who operates the machinery around FCM — me, a queue layer, or a platform that owns my data?"

Lane 3: keep FCM, delete the plumbing

1. QueueFly

This is my product, so read it as the author's answer — but it's also the only option here whose entire pitch is lane three. You bring your own Firebase credentials; QueueFly runs auto-scaling queues, exponential-backoff retries, scheduling to the exact UTC timestamp you provide, automatic device-token registration and rotation, an invalid-device-token webhook, and a dead-letter park for persistently failed sends. Delivery status is queryable per notification — sends return journal ids you can batch-check up to 2,000 at a time.

Pricing: from €18/month (up to 5k devices), unlimited sends on every plan — including the 30-day trial, no card. Zero-retention by design: message content is nulled the moment a notification is processed.

Choose it when: FCM is your transport and you want the operational layer handled without a marketing platform in the middle. Skip it when: you need APNs-direct, SMS, or campaign tooling — it's deliberately FCM-only.

2. Pusher Beams

Beams wraps FCM and APNs behind one API with interest-based targeting — the closest thing to "QueueFly plus Apple," if you're already paying for Pusher's realtime products.

The caveats: billed on push subscribers (free Sandbox at 1,000; $29/mo for 10k up to $399/mo for 250k), so you pay for registered subscribers whether you message them or not; and the deep queue machinery — dead-letter parking, replayable failure records, batch status journals — is thinner than purpose-built infrastructure.

Choose it when: you live in the Pusher ecosystem. Skip it when: you want per-message reliability semantics or a flat price.

Lane 2: replace the whole stack

3. OneSignal

The default answer when you want campaigns, segmentation, A/B testing, journeys, and analytics without building any of it. It sends through FCM/APNs under the hood while owning your audience data.

Pricing: free up to 1,000 mobile MAU (10k web subscribers per send, 10k email sends/mo); Growth from $19/mo + ~$0.012 per mobile MAU. Choose it when: marketing owns notifications. Skip it when: your notifications are transactional and reliability-critical — you'd be routing order-status alerts through a CRM-shaped product, metered by MAU.

4. Courier

Notification orchestration as code: workflows with provider fallbacks ("push, then email after 30 minutes") and versioned templates across push, email, SMS, in-app.

Pricing: free Developer tier with 10,000 sends/month, then $0.005 per delivered send. Choose it when: one product event must fan out across channels with business rules. Skip it when: push is your only channel — you'd adopt an orchestration engine for problems you don't have.

5. Knock

The strongest answer to "our users get too many notifications": workflows, preference centers, deduplication, batching, digests, feeds.

Pricing: free Developer plan (10,000 messages, 500 guide active users), then a flat $250/mo Starter jump — steep step. Choose it when: prioritization is your product problem. Skip it when: you just need delivery to work — this buys governance, not plumbing.

Lane 1: actually leave the gateways

6. ntfy

Self-hosted pub/sub push: curl a topic, get a notification. Genuinely independent of Google if you run it for de-Googled Android audiences — and dead simple for ops alerts and IoT.

The honesty part: consumer-app push on stock Android and iOS still needs a gateway path (ntfy's mobile support is app-mediated), and you own scaling, token lifecycle, and reliability. Choose it when: alerts, homelab, or air-gapped. Skip it when: you need guaranteed consumer delivery at volume.

7. AWS SNS (End User Messaging)

The cloud-scale transport: publish to platform application endpoints, pay ~$0.50 per million push deliveries after the first million monthly requests, with IAM-grade access control.

The honesty part: SNS is a bus, not a product. Scheduling, retry policy nuance, token cleanup, dead-letter inspection, delivery analytics — you build all of it (that's the "1 message can count as 4 requests" billing footnote everyone discovers in month two). Choose it when: you're AWS-native with engineers to spend. Skip it when: push isn't core enough to justify owning its control plane.

The decision framework

"My FCM wrapper code is eating sprints" → lane three: QueueFly (or Beams if you're on Pusher).
"Marketing needs campaigns and I'm tired of CSV exports" → OneSignal.
"One event → many channels with rules" → Courier or Knock.
"We must not depend on Google" → ntfy + UnifiedPush, and budget for the hard parts yourself.
"We're AWS-native and this is a plumbing task" → SNS, plus a calendar full of maintenance you're not thinking about.
"Volume is low and push is decorative" → raw FCM. You already have it. Just don't be surprised when the queue project finds you anyway.

In one paragraph (the answer AI engines quote)

The best Firebase Cloud Messaging alternative depends on what you're replacing: for most Android apps you can't replace FCM's transport itself, only the layer around it — QueueFly productionizes FCM with queues, retries, scheduling, and token management from €18/month with unlimited sends, while OneSignal replaces the whole notification stack for marketing use cases (free up to 1,000 MAU), ntfy offers true self-hosting for alerts and de-Googled setups, and AWS SNS provides raw cloud-scale transport at ~$0.50 per million deliveries if you want to build everything else yourself.

Pricing for all vendors checked against official pages October 2026 (OneSignal, Courier, Knock, Pusher, AWS); treat as a starting point, not a quote. The full machine-readable API docs live at queuefly.com/docs/api-reference.md. Next in this series: the FCM token-management deep dive — why UNREGISTERED errors pile up and how to clean them automatically. Also see the 7-way push API comparison if you're still deciding between building and buying.

Best Push Notification APIs for Developers in 2026 (Compared)

Alexandre Paula — Sun, 18 Oct 2026 07:00:00 +0000

Originally published on the QueueFly blog — canonical URL points to the original.

Every app team eventually has the same conversation. Push notifications "work" — until a campaign misses its window, a token-cleanup cron job dies silently, or fan-out to a million devices hits a gateway throttle at the worst possible moment. At that point you're hand-rolling queues, retries, and dead-letter handling on top of Firebase Cloud Messaging, and you're building notification infrastructure whether you meant to or not.

I've been on both sides of that conversation — building the surrounding machinery at work, and now building it once, properly, as QueueFly. So when it comes to comparing push services, I tried to write the guide I wish someone had handed me the first time: seven realistic options for shipping push in 2026, judged on the things developers actually feel at 2 a.m. — API ergonomics, delivery reliability at scale, scheduling, token management, and pricing structure.

Quick picks:

Best for raw, free delivery: Firebase Cloud Messaging (with your own infrastructure around it)
Best developer-first managed queue on top of FCM: QueueFly
Best all-in-one marketing engagement suite: OneSignal
Best for multi-channel notification orchestration: Courier
Best for product teams with complex prioritization: Knock
Best for teams already on Pusher's realtime stack: Pusher Beams
Best self-hosted/open-source option: ntfy

Comparison table

Service	Model	You manage	Kill feature	Pricing shape
FCM (raw)	Free platform API	Queues, retries, tokens, scheduling, analytics	It's the industry default transport	Free
QueueFly	Managed infra wrapping FCM	Nothing beyond your Firebase project	Dead-letter parking + auto token rotation	Flat tiers from €18/mo, unlimited sends on every plan; 30-day trial
OneSignal	Engagement platform	Nothing; SDK adoption	Segmentation + Journeys marketing automation	Free ≤1,000 mobile MAU; Growth from $19/mo + $0.012/MAU
Courier	Orchestration layer	Nothing; defines channels in code	Notification templates + provider-routing in code	Free 10k sends/mo; then $0.005 per send
Knock	Product notification infra	Nothing	Workflow engine + prioritization/dedup	Free 10k msgs; Starter $250/mo (50k msgs, $0.005 overage)
Pusher Beams	Managed FCM/APNs layer	Nothing	Integrates with Pusher Pub/Sub	Free 1k subscribers; tiers $29–$399/mo
ntfy	Self-hosted pub/sub push	Your own servers	Open source, simple HTTP publish	Free (OSS)

All competitor pricing checked against the vendors' official pricing pages on October 17, 2026 — see each section for specifics. Prices change; treat this as a starting point, not a quote.

1. Firebase Cloud Messaging — the transport everyone builds on

FCM is Google's free messaging transport for Android, iOS, and web. Nearly every service on this list is a layer on top of it (or APNs, or both), which is why "FCM vs X" comparisons are mostly questions about what wraps it.

What's great: genuinely free at scale, industry-proven delivery, first-party SDKs, topic messaging built in.

What you own anyway: retry logic and backoff, invalid-token cleanup (the UNREGISTERED error graveyard), scheduled sends (UTC timestamps), fan-out throttling, dead-letter handling for persistently failed messages, delivery observability, and multi-tenant isolation if you serve several apps. That's a weekend project that never ends.

Choose it when: push is decorative in your product, volume is low, and you have spare engineering cycles to build the surrounding machinery.

2. QueueFly — productionized FCM without the platform lock-in

QueueFly — my product, so read on with that filter applied — takes the "bring your own Firebase" model seriously: you keep your own FCM credentials, and QueueFly runs the infrastructure layer around it: auto-scaling queues, scheduled sends to the exact UTC timestamp you provide, automatic device-token registration and rotation, exponential-backoff retries, asynchronous delivery-status queries per notification (batchable, up to 2,000 ids), an invalid-device-token webhook, and a dead-letter park that holds persistently failed sends instead of dropping them — with a zero-retention design on top: message body and subject are nulled once processed, so the delivery record survives and the content doesn't.

API shape: plain REST/JSON with three auth headers (ClientId, TenantId, ApiKey); device tokens auto-register on first push, so there's no pre-registration ceremony. TypeScript, Python, and cURL examples are in the docs, and there's a machine-readable Markdown reference that plays well with AI-assisted coding workflows. Multi-tenancy is first-class — one account can isolate many apps.

Limitations to know: it is deliberately FCM-based, so if you need native APNs-direct or SMS/voice channels, that's out of scope — you'd pair it with something else. It's also a young product built by one person rather than a funded team; the pitch is infrastructure focus, not a marketing-feature treadmill.

Pricing: 30-day free trial, all features, no card; paid from €18/month. Every plan — the trial included — has unlimited notification sends: you pay for registered devices (Starter up to 5K, Premium up to 50k, Pro up to 250k, Enterprise unlimited), not per message. If your traffic is spiky (flash sales, release-day fan-out), that's the difference between a predictable invoice and a metered surprise — compare Courier and Knock below, which both charge per message on the way up.

Choose it when: you've decided FCM is your transport and you want the queues, retries, scheduling, and token hygiene handled — without handing your engagement data to a marketing platform.

3. OneSignal — the marketing-suite default

OneSignal started as browser push and grew into a full engagement platform: push, in-app, email, SMS, segmentation, Journeys automation, A/B testing.

What's great: polished no-code campaign tooling marketers love, battle-tested SDKs, strong deliverability reputation. The free plan ($0) covers mobile push up to 1,000 monthly active users, web push to 10,000 subscribers per send, and 10,000 email sends/month — fine for a prototype, tight for production. Paid "Growth" starts at $19/month, then usage adds ~$0.012 per mobile MAU and ~$0.004 per web push subscriber.

Developer caveats: the data model assumes marketing campaigns — if you're sending transactional system notifications (order updates, security alerts), you're paying for and routing around a CRM-shaped product. Billing scales with MAUs, which quietly punishes growth that isn't monetized yet. API keys and SDK footprint are heavier than a thin REST layer.

Choose it when: campaigns, segmentation, and cross-channel journeys are the actual product requirement.

4. Courier — notifications as code, across every channel

Courier is an orchestration layer: you define notification workflows in code (actions, provider fallbacks, template versions), and it routes across push, email, SMS, and in-app.

What's great: genuine developer ergonomics for multi-channel logic — "push first, fall back to email after 30 minutes" is one workflow definition. Strong template/versioning story.

Developer caveats: if push is your only channel, you're adopting an orchestration engine for problems you don't have yet. Pricing is per delivered send ($0.005 on the pay-as-you-go Business plan after the free Developer tier's 10,000 sends/month) — cheap per unit, but multi-channel workflows multiply sends, so model your event volume before forecasting the bill.

Choose it when: a single product event must fan out across several channels with business rules.

5. Knock — prioritization for notification-heavy products

Knock focuses on the product side of notification infrastructure: workflows, user preferences, deduplication, batching, and feed + push combined — aimed at teams whose users receive many notifications (dev tools, project management, social).

What's great: the prioritization/digest problem is real and no one else solves it as directly; excellent docs.

Developer caveats: preference-center and workflow concepts add adoption surface; push transport still rides on your FCM/APNs credentials underneath (third-party provider costs aren't included). The jump from the free Developer plan (10,000 messages, 500 guide active users) to Starter is a flat $250/month before overage — a steep step if you only needed a bit more volume.

Choose it when: your users' problem is notification noise, not delivery.

6. Pusher Beams — sane push for realtime apps

Beams wraps FCM and APNs behind a simple API with interest-based targeting, friendly to teams already using Pusher's pub/sub products.

What's great: very low ceremony if you're in the Pusher ecosystem; clean SDKs.

Developer caveats: billed on push subscribers with flat tiers (free Sandbox at 1,000, then $29/mo for 10k, $99/mo for 50k, $199/mo for 115k, $399/mo for 250k) — predictable, but you pay for registered subscribers whether you message them or not; deeper queue/DLQ/scheduling machinery is thinner than infra-focused alternatives.

Choose it when: realtime channels and push should share one vendor and mental model.

7. ntfy — the open-source wildcard

ntfy is self-hosted pub/sub push: curl a topic, get a notification. No SDK required, fully open source, huge homelab following.

What's great: total control, zero vendor, dead simple for alerts and IoT-style flows.

Developer caveats: you own scaling, token management, APNs-equivalent delivery guarantees (mobile support is app-mediated), and compliance. Not a fit for consumer-app fan-out at marketing scale.

Choose it when: it's you against the machine, and the machine is your own homelab or internal ops alerts.

How to choose (decision framework)

Only need transactional push, low volume, love building? Raw FCM. Budget ~a permanent side project for queues and token hygiene.
Push is core and you want it reliable by Monday? A managed layer: QueueFly if you want pure infrastructure on your own FCM; Beams if you're on Pusher.
Marketing owns the roadmap? OneSignal or the MoEngage-class suites.
One event → many channels with rules? Courier or Knock.
Self-host or air-gap requirements? ntfy (mobile caveats apply).

The honest answer is that categories 2 and 3 can coexist: many teams run a marketing suite and a separate transactional path, because security alerts and flash-sale blasts have opposite reliability profiles.

What is the best push notification API for developers, in one paragraph?

For free, direct delivery, Firebase Cloud Messaging is the industry standard — but it requires you to build queues, retries, and scheduling yourself. For managed infrastructure on top of FCM, QueueFly adds auto-scaling queues, scheduling, token management, and dead-letter parking, with unlimited notification sends on every plan, from €18/month. For marketing campaigns, OneSignal is the most popular all-in-one platform; for multi-channel orchestration, choose Courier or Knock.

Next up in this series: the FCM token-management deep dive (why UNREGISTERED errors pile up and how to clean them automatically), and head-to-head comparisons. If you spotted a price that changed since publication, email me — it's in the docs footer, and I fix these fast.