Free plan Managed Kafka for $0 a month. No end date.
Free Kafka, with no card and no time limit
A fully managed, serverless Kafka cluster that carries real production traffic: about 1,000 events a second, more than 80 million a day, every one of them read back twice. Not a trial and not credits, but a plan that stays at $0 for as long as it fits. No credit card, nothing to size, nothing to operate, and the Kafka clients you already use.
- No credit card
- Not time-limited
- About 1,000 events a second
- No fees for public internet traffic
What the free plan includes
$0 / month
- Data in
- 200 KiB/s
- Data out
- 400 KiB/s
- Retention
- 1 day
- Topics
- 10
What 200 KiB/s actually buys you
- ~1,000Events a second, measured on real 1.1 KB JSON events
- 83MEvents a day, every day, at $0
- 2×Every event read back by two consumers
- ~1 MiB/sOf raw JSON, before compression
Measured, not guessed. In our compression test, 50,000 real GitHub events averaging 1.1 KB of JSON each, batched and zstd-compressed the way a normal Kafka producer does, landed at 213 bytes each on the wire. The free plan meters those compressed bytes, so 200 KiB/s carries about 960 of them a second.
Measured after compression
The limits count compressed bytes. That changes everything.
Most free tiers meter the raw size of your records. ntail meters the bytes that actually cross the wire, after your producer has batched and compressed them, on every plan including this one. The same 200 KiB/s carries six times more events with two producer settings changed, and no code change.
| How the producer packs events | Bytes per event, measured | Events a second on the free plan | Events a day |
|---|---|---|---|
| JSON, uncompressed, one event at a timeThe default many teams never change | 1,218 | ~170 | 14.5M |
| JSON, snappy, one event at a time | 745 | ~275 | 23.8M |
| JSON, snappy, batches of 100 | 306 | ~670 | 57.8M |
| JSON, zstd, batches of 100The numbers on this page | 213 | ~960 | 83.1M |
| JSON, zstd, batches of 1,000 | 189 | ~1,080 | 93.6M |
| Avro, zstd, batches of 1,000 | 181 | ~1,130 | 97.6M |
Bytes per event are from our measurement of 50,000 real GitHub events averaging 1,148 bytes of JSON each, at each codec's default level. Events a second are 200 KiB/s divided by those bytes; events a day multiply by 86,400. In the same test a mixed stream of nine event types compressed better still, 7.4× at batches of 100, so a real firehose of varied events fits even more. Your events will compress differently. The post explains the two settings and ships the harness, so you can run it on your own data.
Free plan, not a free trial
Free Kafka that stays free
Most managed Kafka services let you try them for free for a while. The ntail free plan is a plan, so there is no end date to plan around and no bill waiting at the end of it.
| ntail free plan | Typical free trial | |
|---|---|---|
| How long it lasts | No time limit | Until the trial period or the credits run out |
| Credit card | Not needed | Often asked for up front |
| How the limits are measured | After compression, so they stretch 5× or more | Usually on raw, uncompressed bytes |
| When it ends | It doesn't. Upgrade only when you outgrow it | Pay, or move your data somewhere else |
| Public internet traffic | Free, at any volume | Often metered, once you pay |
What it's for
What you can run on free Kafka
More than most teams produce in production. The free plan isn't a sandbox.
-
Your product's event backbone
Signups, orders, payments, notifications and activity feeds, flowing between your services. About 1,000 events a second is more than most growing products push in production, with room left over.
-
Change data capture and pipelines
Stream every change from your database into search, caches and your warehouse, with a full day of history, tens of millions of events, to replay when a consumer needs to catch up.
-
Every environment, production included
The same managed Kafka for development, testing, staging and production, each its own cluster. It's also the easiest way to learn Kafka against the real thing.
When you outgrow it
From free to paid without a migration
The free plan and the paid plan run on the same platform and are metered the same way, so moving up is a change of plan, not a move.
-
Upgrade from the console
Choose the paid plan and add a card. Your clusters, topics and credentials stay exactly where they are.
-
Pay as you go, still measured after compression
Data transfer and storage at the rates on the pricing page, billed per started hour on the same compressed bytes. Estimate your cost before you switch.
-
Public internet traffic stays free
On every plan, sending and reading data over the internet is never charged for, at any volume.
FAQ
Common questions about free Kafka
Is ntail's Kafka really free?
Yes. The free plan costs $0 a month and is never invoiced. It is a real, managed Kafka cluster with the limits listed on this page, not a trial that turns into a bill.
Do I need a credit card to sign up?
No. You sign up with an email address and add a billing address, but no card. A card is only asked for if you choose the paid plan.
Is the free plan time-limited?
No. There is no trial period and no credits that run out. You can stay on the free plan for as long as your workload fits inside its limits.
What are the limits of the free plan?
200 KiB/s of data in and 400 KiB/s of data out, both measured after compression, one day of retention, and up to 10 topics. With typical JSON events batched and compressed, that is about 1,000 events a second written and every one of them read back twice. Public internet traffic is free, as it is on every plan.
Are the limits measured before or after compression?
After. ntail meters the compressed bytes your clients actually send and receive, on the free plan and the paid plan alike. In our measurements, batches of real JSON events compress 5 to 6 times with zstd, so 200 KiB/s on the wire is roughly 1 MiB/s of raw JSON.
How many events a second can I send on the free plan?
About 1,000 a second with typical settings, or more than 80 million a day. We measured 50,000 real GitHub events averaging 1.1 KB of JSON each; batched and zstd-compressed the way a normal Kafka producer does, they took 213 bytes each on the wire, and 200 KiB/s carries about 960 of those a second. Smaller events, or better packing, mean more.
Can I run production workloads on the free plan?
Yes. The free plan runs on the same managed platform as the paid plan, and at about 1,000 events a second it carries more traffic than many products push in production. When you need more throughput, retention or topics, upgrade to the paid plan without moving anything.
What happens when I outgrow the free plan?
Upgrade to the paid plan from the console. It is pay as you go for data transfer and storage, measured after compression just like the free plan, billed per started hour and invoiced monthly. Your clusters stay where they are.
Where is my data stored?
In the region you choose. ntail runs in both the EU and the US: pick the region when you create a cluster and your data is stored and processed there, on infrastructure we operate ourselves.
Start with free Kafka today
About 1,000 events a second, a managed cluster in minutes. No card, no time limit.