Harness

Harness feature flags: plan limits, the FME split, and what the Split acquisition changed

Harness sells feature flags twice: the module it built in 2022, and the Split Software product it bought in 2024. They are different products with different docs and different SDKs, and the roadmap has moved to one of them. If you are evaluating Harness for flags, that fork is the thing to understand first, ahead of any feature comparison.

No card required

The short answer

There are two Harness feature flag products. Harness Feature Flags (FF) is the module Harness built into its own platform in 2022. Harness Feature Management and Experimentation (FME) is Split Software, which Harness agreed to acquire on May 29, 2024 and closed that June. The FF documentation now carries a banner pointing customers using FME to the FME docs for the latest features and SDK updates, and new roadmap items are tracked under FME, though Harness has not published an end-of-life date for FF. Published FF plan limits are Free: 2 developers and 25,000 client MAUs; Team: 50 developers and 100,000 MAUs; Enterprise: unlimited developers and 1 million MAUs, all with unlimited flags and environments. Harness does not publish a price for either product: both paid tiers route to Contact Sales.

Last updated July 2026 · verified against developer.harness.io and harness.io/pricing

01 · Two products, one vendor

Work out which one a Harness rep is selling you

This is not a pedantic distinction. The two products have separate documentation trees, separate SDK families and different underlying concepts, and a comparison article about one tells you almost nothing about the other.

Dimension	Harness Feature Flags (FF)	Harness FME
Origin	Built by Harness, module released 2022	Split Software, acquired 2024
Position today	Documented and supported, no published EOL date	Where new roadmap items are tracked
Docs banner	Directs FME users to the FME docs for the latest features and SDK updates	The current documentation set
Strength	Sits natively inside Harness pipelines and governance	Deep experimentation and statistics, the reason Split had a following
SDKs	Harness FF SDK family	Split SDK family, migrated accounts move to the Harness UI
OpenFeature	Community provider, not on the project's official vendor list	Split provider available
Buy it standalone?	Yes, historically	Yes, sold standalone and bundled in a platform contract

The practical question to put to a rep, in writing: which product am I buying, what is the committed support horizon for it, and if the answer is FF, what is the migration path and who pays for it. A vendor mid-consolidation should be able to answer all three, and the answer belongs in the contract rather than an email.

02 · The plans

Published limits, unpublished prices

Harness documents what each plan includes. It does not document what any of it costs.

Dimension	Free	Team	Enterprise
Developers	Up to 2	Up to 50	Unlimited
Client MAUs included	25,000	100,000, more purchasable	1 million, more purchasable
Feature flags	Unlimited	Unlimited	Unlimited
Environments	Unlimited	Unlimited	Unlimited
Progressive delivery	Manual	Automated	Automated
Governance	Basic	Integrations and support tiers	Policy-based governance, enterprise security, custom dashboards
Support	Community	Standard, 8am to 5pm PST	Premier, 24x7 available
Published price	$0	Contact sales	Contact sales

On the numbers you will find elsewhere

Search for Harness feature flag pricing and you will hit software marketplaces quoting $20 to $50 per 1,000 MAUs for Team, or per-developer platform figures in the $50 to $100 range. We are not going to repeat those as fact, because none of them come from Harness. They are resale and benchmarking estimates, and in a contact-sales product the spread between a listed estimate and a signed price is wide by design. Use them to know roughly what neighborhood you are in before the first call, then get a written quote.

The two-year-old free tier limit worth flagging: 2 developers is small. It is enough to prototype and not enough to run a team pilot, which is presumably the intent. Compare that to the free tiers of the vendors in our best feature flag tools roundup, several of which allow unlimited seats on free plans.

03 · The metering unit

MAU pricing rewards some products and punishes others

Harness meters client monthly active users. Understanding what that favors is more useful than a feature checklist, because the meter, not the feature list, is what decides whether a tool is affordable for your specific shape of product.

MAU metering suits you if

You sell B2B software to a few hundred or few thousand named users, each of whom is worth real money. A 100,000 MAU allowance is then effectively unlimited, you never think about the meter, and the price behaves like a flat fee. This is the case Harness, LaunchDarkly and most MAU-metered vendors are built for, and in that case it works well.

MAU metering hurts you if

You run a consumer product, a marketplace, a media site, or anything with a large logged-out or anonymous audience. Every anonymous visitor your client SDK evaluates a flag for is potentially an identifier against the meter. Traffic spikes, bot traffic and a viral week all translate directly into cost, and the flag budget becomes a function of marketing performance, which is a genuinely awkward conversation to have.

Compare to request metering

PostHog bills requests to the flags endpoint rather than users, which flips the incentive: you optimize evaluation patterns instead of audience size. Vercel bills flag requests but batches all flags on a page into one. AWS AppConfig bills configuration fetches. Each meter has a shape of product it is cheap for, and the only way to know yours is to model your own numbers against each.

Compare to flat pricing

A flat per-team price removes the forecasting exercise entirely. That is the trade we made in our own feature flag software pricing: unlimited seats and unlimited flags, so nobody has to predict next quarter's MAU curve to know next quarter's bill. It is not free of trade-offs, but the trade-off is legible.

04 · The walls

Four things to weigh before committing

Wall 1

You are buying during a product consolidation

Two feature management products under one roof is a transitional state, not a destination. Harness has been clear about where new work goes. What it has not published is a support horizon for the original module, which means the risk of a migration lands on the buyer without a date to plan against. That is manageable if you ask for commitments in the contract, and a real risk if you do not.

Wall 2

No published price at all

Every paid tier is Contact Sales. That is normal for enterprise DevOps platforms and it has a specific cost: you cannot compare Harness to anything without entering a sales cycle, and you cannot forecast a renewal from public information. Vendors that publish a rate card, even a usage-based one, let you do arithmetic on a Tuesday afternoon. Vendors that do not are asking for a meeting first.

Wall 3

The platform is the pitch

Harness sells a full DevOps platform: CI, CD and GitOps, feature management, cloud cost management, security testing orchestration, chaos engineering, more. If you want all of it, the bundling is genuine value and one vendor relationship is easier than six. If you want flags, the surrounding platform is scope you are paying for in complexity even where you are not paying in dollars: an onboarding, a governance model, and a set of concepts sized for something much larger.

Wall 4

OpenFeature status is community, not official

On the OpenFeature project's own vendor dataset, the Harness provider is community-maintained rather than officially vendor-supported, unlike Split, ConfigCat, Flagsmith, DevCycle, Datadog and others. If your abstraction strategy is standardizing on OpenFeature so you can swap vendors later, check that the provider for whichever Harness product you buy is maintained at the level you need. Our take on whether that abstraction earns its keep is in the OpenFeature adoption analysis.

05 · The honest fork

Should you buy Harness for flags? It hinges on one question.

Use Harness

It makes sense if all of this is true

You already run Harness for CI, CD or cost management, so feature management is a line on an existing contract rather than a new vendor, a new security review and a new procurement cycle.
You want experimentation with a serious statistics engine behind it, which is what the Split acquisition actually bought.
Your product is B2B with a bounded, identified user base, so MAU metering behaves like a flat fee.
You need policy-based governance and approval workflows on flag changes, which is genuinely where enterprise DevOps platforms are strong.
You have the procurement patience for a contact-sales motion and the leverage to negotiate.

Said plainly: if Harness is already your platform, buying flags anywhere else is the harder argument. Ask for FME rather than FF, and get the support horizon in writing.

Look elsewhere

Look further when any of this is true

You are not a Harness customer and do not intend to become one. Adopting a platform to get one capability is a large tail of cost.
You need a price today to build a business case, not after two discovery calls.
You run a consumer or high-traffic product where MAU metering scales with your marketing rather than your value.
You would rather not adopt a product whose successor is still being consolidated.
You want unlimited seats so support, product and design can flip flags without a per-developer line item.

A focused feature flags platform with a published flat price, unlimited seats and unlimited flags does one job. See the LaunchDarkly alternative comparison for the same argument against the category leader, or best feature flag tools for the full field with verified pricing.

Curious where else the Split engine turns up? Microsoft states that experimentation in Azure feature flags is powered by Split Software running on Azure, so the same statistics engine reaches buyers through two very different front doors.

06 · FAQ

Questions people actually search

What are Harness Feature Flags?

Harness Feature Flags is the feature management module Harness launched in 2022 as part of its DevOps platform, sitting alongside CI, CD, Cloud Cost Management and the rest. It gives you flags, targeting, percentage rollouts and a relay proxy, with the pitch being that flags live in the same platform as your pipelines rather than in a separate vendor.

Is Harness Feature Flags being replaced by FME?

Not formally retired, but new development has moved. Harness acquired Split Software in 2024 and rebranded it Feature Management and Experimentation. The Feature Flags docs now carry a banner directing customers who have migrated to or are using FME to the FME documentation for the latest features and SDK updates, and new roadmap items are tracked under FME. Harness has not published an end-of-life date for the original module.

What is the difference between Harness FF and Harness FME?

FF is the 2022-era module built inside the Harness platform. FME is the former Split Software product, acquired in 2024, with a much deeper experimentation and statistics engine behind it. They are two different codebases with two documentation sets, two SDK families and different concepts. FME is where the investment is going.

How much do Harness Feature Flags cost?

Harness does not publish a list price for feature management. The plans page routes to Contact Sales for both paid tiers. Third-party marketplaces report figures in the range of $20 to $50 per 1,000 monthly active users for the Team tier, but those are resale estimates rather than vendor-published prices, so treat them as a starting point for negotiation and not a quote.

Does Harness have a free feature flag plan?

Yes. The Free plan covers up to 2 developers and 25,000 client monthly active users, with unlimited flags and unlimited environments. The Team plan raises that to 50 developers and 100,000 MAUs included, and Enterprise to unlimited developers with 1 million MAUs included. Extra MAUs are purchasable on both paid tiers.

What is a monthly active user in Harness feature flags?

A client MAU is a unique identifier your client-side SDKs evaluate flags against in a billing month. It is the metering unit for the plan limits, so it counts people rather than requests, seats or flags. That distinction matters: a consumer app with a large logged-out audience can burn through MAUs quickly while a B2B tool with a hundred customers barely registers.

What is the Harness Feature Flags relay proxy?

The relay proxy is a service you run inside your own network that connects to Harness once and serves flag evaluations to your SDKs locally. It is the answer for air-gapped environments, for reducing outbound connections at scale, and for keeping evaluation latency low in a region far from the control plane. Most enterprise feature flag vendors ship an equivalent.

Is Harness a good choice if I only want feature flags?

It depends on whether you already buy Harness. If your CI, CD and cost management already run there, adding feature management is a contract line rather than a new vendor, and that has real value. If you want flags alone, you are buying into a platform sales motion, an unpublished price and an active product transition, which is a lot of overhead for one capability.

Early access · launching soon

A published price, unlimited seats, one job done well

Flags, targeting, rollouts, kill switches and A/B testing on a flat per-team price you can read on the website. Join the early-access list, no card required.

No card required · your data stays yours

Keep exploring