LaunchDarkly pricing explained: the two meters that decide your bill

LaunchDarkly is free forever up to 1,000 client-side MAU and 5 service connections, on a tier called Developer that also includes unlimited flags and unlimited seats.

Past those limits, Foundation is pay-as-you-go, billed yearly after a 14-day trial: $8.33/mo per 1,000 client-side MAU, $10/mo per service connection beyond the included five, and $5 per additional 1,000 AI runs. Enterprise is custom quoted.

There is no per-seat charge on any tier, and no single monthly number on any paid tier. That second fact is why this question is hard to answer and why this page exists.

Figures checked against launchdarkly.com/pricing on 8 September 2026.

The most useful thing to understand about the bill is that LaunchDarkly meters two different things, and which one charges you is decided by where your flags are evaluated, not by how big your product is.

• Client-side MAU meters people. Browser, mobile and edge SDKs.
• Service connections meter processes. Server-side SDKs.

A consumer web app pays for users. A backend service pays for running instances. A team with both pays on both, separately. Almost every estimate that goes wrong does so by assuming MAU is the only meter, or by assuming it counts what an analytics tool would call a monthly active user. It does not.

LaunchDarkly's docs define it as “the number of unique monthly active contexts, typically user contexts, evaluated by LaunchDarkly client-side, AI, and edge SDKs.” Three details matter:

• Counting is by context key. The same key evaluated a million times in a month is one MAU.
• Counting is monthly, over unique keys in that month.
• The key is the identity, and changing it mints a new one.

That third point is the expensive one. Their docs state that if you change an existing context key, LaunchDarkly considers the new key unique and counts it as an additional MAU. So if you evaluate flags for logged-out visitors using a per-session or per-device id, every new session creates a new key and therefore a new MAU.

Your billed MAU can be several times your real user count with nothing broken. If you are near a boundary, check how your anonymous contexts are keyed before you check anything else.

Server-side evaluations do not count toward client-side MAU at all. LaunchDarkly's docs say server-side SDKs use service connections as the billing metric, and define one as “one instance of one server-side SDK connected to one LaunchDarkly environment.”

Read that as instances times environments. One service on three replicas talking to production and staging is not the same number as one laptop. Five connections are included on both Developer and Foundation; each additional one is $10 a month.

If you autoscale, model this meter before you commit. It is also worth asking LaunchDarkly directly how concurrency is sampled for your topology, because the public pricing page does not say, and a fleet of short-lived instances is exactly the case where the answer changes the bill.

Decide which SDK your flags run in, then count the matching thing.

Two worked examples. A server-rendered app with two services on two replicas each, across production and staging, is eight connections: three over the included five, so $30 a month, and zero MAU. A consumer app with 10,000 unique browser context keys a month is roughly $83.30 a month at the listed rate.

One caveat worth stating: the public pricing page does not spell out whether Foundation includes the first 1,000 MAU before metering starts. Treat any estimate near that boundary as approximate, and ask them.

It is cheap, often free, when your flags are evaluated server-side in a small number of long-lived instances, or when your browser audience is genuinely under a thousand identified users a month. It is also good value when you want experimentation, session replay and error tracking, because those are bundled rather than bought separately.

It gets expensive with a large anonymous browser audience, because that is exactly the case where the meter counts keys rather than customers, and when you autoscale many short-lived server instances. And it is expensive in the ordinary way if you need approvals, scheduling, custom roles or SCIM, since those sit on a custom-quoted Enterprise tier.

If your conclusion is that you are paying a usage meter to run one if statement, the alternative is a flat price. LaunchFlag is $12 a month for Pro and $29 for Launch, per account: no per-seat charge, no MAU meter, no per-evaluation charge. Flags are evaluated inside your own process, so your traffic never reaches us and there is nothing to meter.

That trade only makes sense if the platform features are ones you do not need. There is no experimentation, no session replay, no approvals, no scheduling and no SCIM. If you need any of them, LaunchDarkly is the right tool and the meter is the price of it. The longer version of that comparison is in LaunchDarkly alternatives.

Prices and tier contents from launchdarkly.com/pricing, checked 8 September 2026. The MAU and service connection definitions are quoted from LaunchDarkly's own billing documentation at calculating billing, checked the same day.

Usage pricing changes without notice and every figure here is in US dollars. Check the source before you rely on any number on this page.