Pricing analysis

LaunchDarkly pricing in 2026: what it actually costs

LaunchDarkly stopped charging per seat. It now meters service connections and client-side monthly active users, which means your bill tracks your pod count and your traffic instead of your headcount. Here are the published rates, the arithmetic, and where it bites.

No card required

The short answer

LaunchDarkly has a free Developer plan at $0 with unlimited seats and unlimited flags. Above that, the Foundation plan is usage-priced and billed yearly: $10 per service connection per month, $8.33 per 1,000 client-side MAU per month, and $5 per 1,000 AI runs past 5,000 per month. Enterprise and Guardian are custom quotes from sales. LaunchDarkly is no longer a per-seat product, so the number that moves your invoice is how many server SDK instances you run and how many client-side users you serve, not how many engineers you hire.

Rates verified on launchdarkly.com/pricing, July 2026 · always check the vendor page, this space changes fast

01 · The plans

LaunchDarkly's four tiers, and what each one actually meters

Two tiers have published prices and two do not. That split matters: once you cross into Enterprise or Guardian, the price is whatever your negotiation produces, and nobody on the internet can tell you what you will pay.

Tier	Published price	What you get	Best fit
Developer	$0	Unlimited seats, unlimited flags, 1 project, 3 environments, 10M logs and traces, 5,000 session replays and errors, 5,000 AI runs a month, 14-day retention. Monthly service connection and MAU entitlements apply, but LaunchDarkly does not publish the figures.	Small teams, side projects, a serious evaluation
Foundation	Usage-priced, billed yearly (rates below)	Pay for what you run and serve: server-side service connections, client-side MAU, and AI runs above the included 5,000. There is no published minimum and no published platform fee. It is pure usage.	Growing teams that outgrew the free entitlements
Enterprise	Custom, contact sales	The governance tier: no public rate card, priced by negotiation against your scale and contract terms.	Regulated, audited, or large multi-team orgs
Guardian	Custom, contact sales	LaunchDarkly's newest tier: release monitoring, guardrail metrics, and automatic rollback. Also sales-led with no public pricing.	Teams that want releases watched and reverted automatically

Foundation rate card, verbatim, all billed yearly

$10

per service connection / month

One instance of one server-side SDK, connected to one environment.

$8.33

per 1,000 client-side MAU / month

Client-side and edge SDKs are billed by monthly active users instead.

per 1,000 AI runs past 5,000 / month

AgentControl usage, above the 5,000 runs included each month.

Those three lines are the entire published rate card. LaunchDarkly does not publish month-to-month (non-annual) rates, experimentation MAU rates, session replay add-on rates, or the free plan's service connection and MAU caps, so we are not going to guess at them here. If you see those numbers quoted elsewhere, ask where they came from.

02 · The two meters

Service connections and client-side MAU

Everything on the Foundation bill comes down to which side of your stack the SDK lives on. Server-side SDKs are counted as connections. Client-side and edge SDKs are counted as people. Get that split wrong in your forecast and you will be off by an order of magnitude in either direction.

Here are LaunchDarkly's own definitions, quoted from its docs so you can hold us to them:

"A service connection is one instance of one server-side SDK connected to one LaunchDarkly environment for a time period measured in minutes."

"Client-side and edge SDKs are billed by monthly active users (MAU) instead of by service connections."

Read the first one twice. Per instance. Per environment. Measured in minutes. A single service running twelve pods in production is twelve service connections. Run the same service in staging with the same SDK pointed at a second environment, and that is counted separately. Autoscale to thirty pods on Black Friday and the meter follows you up, prorated by the minutes those pods were alive.

MAU is the other half. LaunchDarkly counts the contexts that encounter your client-side and edge SDKs, and you pay for your single most-used context kind. If your product is a consumer app, this line will dwarf your connection line, and it moves with a marketing campaign rather than with anything your engineering team decided.

Which meter am I on?

Go / Java / Node / Python / Ruby / .NET server SDK Service connection, per instance
Each running pod or container of that service Another service connection
The same SDK pointed at a second environment Counted separately
Browser JS / React / iOS / Android SDK Client-side MAU
Edge / worker SDK MAU, not connections
AgentControl AI runs above 5,000 a month $5 per 1,000 runs

The practical consequence: a Kubernetes deployment with an aggressive HPA is a pricing decision now, not just an infrastructure one. The same is true of any gradual rollout that ramps traffic on the client side.

03 · Worked examples

What that adds up to, with the arithmetic shown

Everything below is our calculation from LaunchDarkly's published Foundation rates, billed yearly. We are showing the multiplication so you can check it against your own instance count and traffic. These are not quotes from LaunchDarkly, and your real invoice depends on your fleet, your context kinds, and anything you negotiate.

Backend-only team

12 server SDK instances, no client-side SDK.

12 service connections x $10 $120
Client-side MAU $0
AI runs $0

$120 / month

$1,440 a year. Cheap, right up until the pod count moves. Double the fleet to 24 and it is $240 a month for exactly the same flags, the same team, and the same features.

B2C app

200,000 client-side MAU plus 20 server instances.

200 x 1,000 MAU x $8.33 $1,666
20 service connections x $10 $200
AI runs $0

$1,866 / month

$22,392 a year. The MAU line is 89% of it. A successful launch campaign that doubles your active users roughly doubles this bill, and no engineer approved that spend.

Mixed platform team

40 server instances, 60,000 client-side MAU, 25,000 AI runs.

40 service connections x $10 $400
60 x 1,000 MAU x $8.33 $499.80
20,000 AI runs past 5,000: 20 x $5 $100

$999.80 / month

About $11,997.60 a year. Three separate meters, three separate ways for the number to move without anyone changing a flag.

The line to remember

Your LaunchDarkly bill scales with your pod count and your traffic, not your team size.

That is genuinely better than per-seat for a big team running a small footprint. It is genuinely worse for a small team running a big footprint. And it is harder to forecast for everyone, because the two things that drive it (autoscaling and traffic spikes) are the two things you least control. A flat plan is not cheaper by magic. It is just a number you can put in a budget in January and still recognize in December.

04 · What changed

Most of what you'll read about LaunchDarkly pricing is stale

For years, LaunchDarkly was the canonical per-seat feature flag platform. Every comparison post, every Reddit thread, every internal build-versus-buy doc argued about the seat price. That framing is now wrong. Seats are unlimited, including on the free plan, and the rate card does not contain a per-seat number at all.

We are pointing this out even though the old story flattered us. Our own LaunchDarkly alternative page has argued the per-seat case for a long time, and the honest position today is that the seat argument is dead and the forecasting argument is the real one. If a comparison page tells you LaunchDarkly charges $X per seat, it has not been updated and you should distrust the rest of its numbers too.

The shift also tells you where the product is going. AgentControl runs and the Guardian tier (release monitoring, guardrail metrics, automatic rollback) are new surface area. LaunchDarkly is pricing the platform like infrastructure now, which is a coherent bet. It just moves the budgeting risk onto you.

Then vs now

The old model

Per seat

Hiring raised the bill. Teams rationed logins and shared accounts to stay under budget.

The current model

Per connection + MAU

Scaling raises the bill. Your SRE team's autoscaler is now a line item on the finance forecast.

Credit where it is due: unlimited seats on a free tier is a real gift to a small team, and plenty of vendors (Unleash, for one) still charge per person. If you have five engineers and three pods, LaunchDarkly's free plan may genuinely be the best deal in this market.

05 · The wider market

Every vendor meters something different

A bare "$ per month" column across these vendors would be misleading, so we have kept the unit visible in every cell. One charges per seat, one per request, one per event, one per connection plus MAU, and several charge a flat fee. Two of them publish nothing at all. Comparing the headline numbers without the units is how teams end up with a surprise invoice.

Vendor	What it meters	Free tier	Paid entry point	Notes
LaunchDarkly	Per service connection + per client-side MAU	Developer $0, unlimited seats and flags	Foundation: $10 per connection/mo + $8.33 per 1,000 MAU/mo (yearly)	No per-seat charge. No published minimum. Enterprise and Guardian are custom.
Statsig	Per event (flag checks are free)	Developer $0	Pro $150/mo flat (one project) with a metered-event allowance	Then $0.05 per 1,000 events. Enterprise custom.
ConfigCat	Flat per month, real meter is config-download quota	Free $0 (10 flags, 2 environments)	Pro $110/mo flat	Unlimited seats and unlimited MAU on all tiers. Low tiers cap flags and environments. Smart $325, Enterprise $900, Dedicated $4,500.
Flagsmith	Per API request + per seat above the included members	Free $0 (50K requests/mo, 1 seat)	Start-Up $40/mo billed yearly ($45 monthly): 1M requests, 3 members	Scale-Up $250/mo yearly ($300 monthly): 5M+ requests, 5 members. Extra seats $50 to $60 per member/mo. Overages from $50 per million requests.
Unleash	Per seat (genuinely)	Not on the paid plan	Pay-As-You-Go $75 per seat/mo, 5-seat minimum self-hosted (about $375/mo)	Includes 53M API requests, then $5 per million. OSS editions (Unleash OSS, Unleash Edge OSS) are deprecated: LTS from Dec 10 2025, end-of-life Dec 31 2026.
PostHog	Per request, above a free allowance	Feature flags free to 1M requests/mo	Usage: $0.000100 per request from 1M to 2M, cheaper above	No per-seat charge for flags.
Harness FME (was Split.io)	Not published	Not published	Contact sales	split.io/pricing now redirects to harness.io/pricing. No public rate card.
Optimizely	Not published	Not published	Contact sales / request a demo	No public pricing anywhere.
Toggled (us)	Flat per team. Unlimited seats, unlimited flags	No free plan. Early access is the free way in	Launch $49/mo (planned), Team $149/mo, Scale $499/mo	PLANNED launch pricing. We are in early access and not yet launched. Enterprise custom.

Prices verified July 2026. Check the vendor's own page before you sign anything, because this space changes fast: LaunchDarkly's move off per-seat is proof of that. Our own numbers are on the Toggled pricing page, and they are planned launch prices, not live billing.

06 · The honest call

When to pay LaunchDarkly, and when a flat plan wins

Pick LaunchDarkly when

You need enterprise governance: approval workflows, granular RBAC, audit trails that survive a compliance review.
You run serious experimentation and want it from the same vendor as the flags.
Your SDK matrix is exotic. LaunchDarkly supports more platforms than anyone, and that is not a marketing claim, it is a maintenance burden they have actually carried for a decade.
You are proving something at real scale and need a vendor with a decade of production evidence behind it.
You are a small team with a small footprint. The free Developer plan gives you unlimited seats and unlimited flags, which is a genuinely generous offer and hard to beat at $0.

A flat plan is easier when

You autoscale hard. A meter that counts running instances turns a capacity decision into a finance decision.
Your traffic is spiky or seasonal, and a good month should not produce a bad invoice.
You want one number in the budget. Toggled is planned to launch at $49, $149, and $499 a month, flat, with unlimited seats and unlimited flags.
You want A/B testing included rather than metered separately.
You mostly need the core workflow: flags, targeting, rollouts, and a kill switch that works in seconds.

To be plain about our position: Toggled is in early access and has not launched yet. Those prices are planned launch pricing. We have no customer logos to show you and we are not going to invent any. What we do have is a feature flags platform built around feature toggle management, A/B testing, and a kill switch, priced so that scaling your infrastructure never touches the invoice.

07 · Questions people actually ask

LaunchDarkly pricing FAQ

How much does LaunchDarkly cost?

LaunchDarkly costs $0 on the Developer plan. On Foundation you pay usage: $10 per service connection per month, $8.33 per 1,000 client-side MAU per month, and $5 per 1,000 AI runs past 5,000, all billed yearly. Enterprise and Guardian are quoted by sales, with no published rates.

Is LaunchDarkly free?

Yes, partly. The Developer plan is $0 and includes unlimited seats and unlimited flags, 1 project, 3 environments, 10M logs and traces, 5,000 session replays and errors, 5,000 AI runs a month, and 14-day retention. It has monthly service connection and MAU entitlements, but LaunchDarkly does not publish those limits.

Does LaunchDarkly charge per seat?

No, not anymore. LaunchDarkly has moved off per-seat pricing, and seats are unlimited even on the free Developer plan. The meter is now infrastructure and traffic: server-side SDK instances (service connections) plus client-side monthly active users. Any article quoting a per-seat LaunchDarkly price is out of date.

What is a service connection in LaunchDarkly?

LaunchDarkly defines a service connection as "one instance of one server-side SDK connected to one LaunchDarkly environment for a time period measured in minutes." In practice that is each running pod, container, or process holding a server SDK client, counted per environment. Foundation bills $10 per service connection per month.

What is a LaunchDarkly MAU?

A LaunchDarkly MAU is a monthly active user seen by your client-side and edge SDKs. Those SDKs are billed by MAU rather than by service connection, counting the contexts that encounter them, and you pay for your single most-used context kind. Foundation charges $8.33 per 1,000 client-side MAU per month.

Is LaunchDarkly worth it?

It depends on your shape. If you need deep governance, approvals, audit trails, mature experimentation, and a very broad SDK matrix at enterprise scale, LaunchDarkly earns its price. If you are a small team with a large pod fleet or heavy consumer traffic, usage pricing can outrun the value you get from it.

What is the cheapest LaunchDarkly alternative?

The cheapest option is usually a free tier: PostHog, ConfigCat, Flagsmith, and Statsig all have $0 plans, and LaunchDarkly's own Developer plan is free with unlimited seats. Among paid plans, flat pricing is the cheapest to forecast. Toggled's planned launch pricing starts at $49 a month with unlimited seats.

Last updated July 2026 · rates verified against launchdarkly.com/pricing and LaunchDarkly's docs

Early access · launching soon

A bill you can actually forecast

Join the early-access list and lock in flat launch pricing: unlimited seats, unlimited flags, no connection meter. No card required.

No card required · your data stays yours

Keep exploring