Flagsmith Alternatives
Seven replacements compared by what you pay for: requests, seats, monthly active users or server connections.
Flagsmith is a solid open source feature flag tool, but its cloud bill follows API requests, governance starts on its $300 Scale-Up plan, and its analytics stay light. We compared seven alternatives on the unit each one bills, its licence, self-hosting and the plan that unlocks approvals. PostHog is our first pick if you want analytics and 1,000,000 free flag requests a month. GrowthBook is the open source move with real experiments, Unleash removes per-request billing and adds approvals on a self-serve plan, and ConfigCat gives a flat price with unlimited seats. LaunchDarkly, Statsig and DevCycle complete the list for enterprise governance, free flag checks and OpenFeature.
The ranking is editorial: a sponsored placement is always flagged on its card, and no affiliate link changes a score.
Why teams leave Flagsmith
Flagsmith does its core job well. It serves feature flags and typed remote config values to web, mobile and server apps, the bulk of its code is BSD-3-Clause, and its free self-hosted build has no API limits. The reasons people move are about what sits around the flags: the way requests are billed, which plan unlocks governance, and how far the analytics go.
Requests are the bill, and they climb with traffic
Free covers 50,000 requests a month, Start-Up 1,000,000. Each client SDK initialisation is one request, and a server-rendered app makes one per page view. Past the quota, the Start-Up overage rate is $7 per 100,000 requests, which is $70 per million. A Start-Up app at 5 million requests pays its $45 plan and then about $280 of overage. A first overrun within 200% gets a 30-day grace period; repeat it and the bill arrives within 30 days.
Governance starts at $300 a month
Roles, permissions, change requests, audit logs and SAML SSO all start on Scale-Up, at $300 a month billed monthly, a little less on yearly billing. Start-Up gives 2FA and email support. A team that only needs four-eyes approval on production flags jumps from $45 to $300 a month, while Unleash includes change requests and SSO on its self-serve plan.
Team members are counted and capped
Free has 1 team member, Start-Up 3. Scale-Up includes 5 and stops at 20, each extra one at $60 a month on monthly billing. Twelve people on Scale-Up cost $720 a month billed monthly.
Open source covers the flags, not the governance
Self-hosting the open source build is free, but enterprise-level governance and management features need a valid Flagsmith Enterprise licence. The supported self-hosted and private cloud editions sit on Enterprise, priced through sales.
Experiments need analysis elsewhere
Flagsmith runs A/B and multivariate tests from Start-Up and lets you bring your own experimentation warehouse on Scale-Up, but its plan list names no CUPED or sequential testing, the statistics GrowthBook and Statsig ship. One reviewer in 2025 asked for more built-in analytics and experimentation; another found that feature-specific segments count towards the total segment limit and had to merge segments to stay under it.
The export leaves your targeting behind
Flagsmith's bulk export carries flags, flag states and multivariate weights. It does not carry feature-based segments, segment overrides, custom fields, schedules, tags or owners. Plan to rebuild your targeting rules in the new tool, and treat the export as a starting point.
PostHog
PostHog is a product analytics suite where feature flags sit next to events, funnels, session replay and experiments. You create a flag, roll it out to 10% of users, and read its effect on the same dashboards, with no warehouse and no second vendor.
PostHog gives 1,000,000 flag requests every month for free, twenty times Flagsmith's free 50,000, and the free plan has unlimited team members where Flagsmith Free has one. Past the free tier, pricing starts at $0.0001 per request and drops with volume. That is $100 per million at the top rate, against $70 per million on Flagsmith's Start-Up overage, so a very high-traffic app with client-side flags can cost more here. The fix is local evaluation on your servers: the SDK polls flag definitions instead of asking for every user, and a poll counts as 10 requests when the definitions changed and 1 when they did not.
Its limit is hosting. The code is MIT outside the ee/ directory, but self-hosted deployments are officially unsupported and ship without tagged releases. If you self-host Flagsmith for compliance, PostHog Cloud is a change of model, not a like-for-like swap. Governance is sold separately as platform packages: SAML SSO comes with Scale at $750 a month and role-based access control with Enterprise, where Flagsmith puts SAML on its $300 Scale-Up plan. And if PostHog itself is the tool you outgrow later, our PostHog alternatives cover the next step.
- 1,000,000 flag requests free every month
- Flags, experiments, analytics and replay share one SDK
- Billing limits that stop the bill instead of surprising you
- ✓Twenty times Flagsmith's free request quota
- ✓No per-seat fee on the free plan
- ✓Experiment results read on the same product data
- ✗Top-rate requests cost $100 per million, above Flagsmith Start-Up's $70 overage
- ✗Self-hosting is officially unsupported
- ✗A lot of platform if you only need on/off flags
| Criterion | PostHog | Flagsmith |
|---|---|---|
| Free requests a month | 1,000,000 | 50,000 |
| Free plan seats | Unlimited | 1 team member |
| Built-in analytics | Yes, full product analytics | A/B and MVT on Start-Up |
| Self-hosting | Possible, officially unsupported | Open source or Enterprise |
Pick PostHog when the reason you leave Flagsmith is the missing analytics around your flags and your traffic fits a usage bill. If you must run everything on your own servers with vendor support, look at GrowthBook or Unleash instead.
GrowthBook
GrowthBook is an open core platform for feature flags, experiments and product analytics. Its code is MIT licensed outside a few enterprise directories, the free self-hosted plan has unlimited users, and even the free cloud plan has unlimited flags, experiments and traffic. It reads experiment results from your own data warehouse, or from a managed warehouse GrowthBook runs for you, so your event data can stay where it already lives.
For a Flagsmith user this is the closest open source move that also upgrades experimentation. Flagsmith gives you A/B and multivariate flags, but GrowthBook adds a full statistics engine: sequential testing, CUPED variance reduction and multi-arm bandits on Pro. On the free self-hosted edition you keep the Bayesian and frequentist engines, but sequential testing, CUPED and bandits move to Enterprise, and the free edition allows 1 project. Billing is by seat, not by request or monthly active user; traffic only counts past Pro's allowances: 2 million CDN requests a month ($10 per extra million), 20 GB of CDN bandwidth ($1 per extra GB) and 2 million events each on the managed warehouse and the event forwarder ($30 per extra million).
The catch is the cloud Pro plan. At $40 per seat per month it is capped at 30 users and 3 projects, and CDN requests above 2 million a month cost $10 per extra million. A team of 12 pays $480 a month on Pro, well under Flagsmith Scale-Up with 12 members ($720 billed monthly). Approval workflows, custom SSO beyond Google login and SCIM sit on Enterprise, priced through sales.
- MIT core, free self-hosting with unlimited users
- Warehouse-native experiment analysis
- Seat pricing, with CDN overage only past 2M requests
- ✓Self-hosted plan free with no user cap
- ✓Sequential testing, CUPED and bandits on Pro
- ✓Local evaluation and a self-hosted proxy for SDKs
- ✗Pro is capped at 30 users and 3 projects
- ✗Approval workflows and custom SSO need Enterprise
- ✗Free self-hosted: 1 project, no CUPED or sequential tests
| Criterion | GrowthBook | Flagsmith |
|---|---|---|
| Licence | MIT, enterprise directories excluded | BSD-3-Clause, enterprise features closed |
| Billing unit | Seats | API requests and team members |
| 12 people, cloud, monthly | $480/mo (Pro) | $720/mo (Scale-Up) |
| Approval workflows | Enterprise | Scale-Up |
Pick GrowthBook if you self-host and want experiments measured on your own data under an MIT core. If you need change requests on a priced plan, or more than 30 users without an Enterprise contract, Unleash fits better.
Unleash
Unleash is an open source feature flag server with 14.3k stars on GitHub. The server code is AGPL-3.0 since version 8.0.0 and was Apache 2.0 before, so comparisons that call it Apache-licensed describe older versions. AGPL matters if you modify the server and serve it to others. Its pricing flips Flagsmith's model: you pay $75 per seat per month on Pay-As-You-Go, and client-side MAU and service connections are unlimited.
That plan already carries what Flagsmith keeps for Scale-Up and Enterprise: four-eyes approvals, change requests, SSO with SAML 2.0 or OpenID Connect, custom project roles and audit logs kept up to 2 years. On Cloud, 53 million API requests a month are included, then $5 per million, where Flagsmith's Start-Up overage works out at $70 per million. Those requests come from your SDKs syncing, not from your users: Unleash's own example, 200 backend instances polling at the default 15 seconds, makes roughly 34.6 million a month. Server-side SDKs evaluate flags locally, so a busy backend does not burn requests.
The limits are the open source edition and the seat floor. Unleash OSS allows 1 project and 2 environments, with no SSO and no change requests. There is no free cloud plan, and self-hosting the paid edition needs at least 5 seats, so $375 a month is the floor.
- Unlimited MAU and service connections
- Change requests and SSO on the self-serve plan
- 53M API requests a month included on Cloud
- ✓Traffic growth does not change the bill
- ✓Governance without an Enterprise contract
- ✓Export and import built in for migration
- ✗Open source edition: 1 project, 2 environments
- ✗5-seat minimum self-hosted, so $375 a month at least
- ✗No CUPED or sequential tests on its plan list
| Criterion | Unleash | Flagsmith |
|---|---|---|
| Licence | AGPL-3.0 | BSD-3-Clause |
| Billing unit | Seats | API requests and team members |
| Overage on cloud | $5 per million past 53M | $7 per 100,000 on Start-Up |
| Change requests | Pay-As-You-Go | Scale-Up |
Pick Unleash when requests are what make your Flagsmith bill grow, or when you need approvals and SSO on a self-serve plan. For a team of two or three on a tight budget, the 5-seat floor of $375 a month costs more than ConfigCat Pro or PostHog's free tier.
ConfigCat
ConfigCat is a hosted service for feature flags and remote configuration, the two jobs at the heart of Flagsmith. Every plan has unlimited seats, unlimited MAU and unlimited flag reads. You pay a flat fee for a plan: Pro is $110 a month billed annually, or $120 billed monthly, for 100 flags, 3 environments and 25 million config downloads a month.
The unit is what makes it predictable. The SDK downloads a config file from the CDN and caches it, and you can evaluate flags as many times as you like between downloads: it still counts as one. The polling interval sets the count: ConfigCat's own example, a web app with 15,000 users who spend 5 hours a month in it, polling every 60 seconds by default, makes 4.5 million downloads a month, close to the Free plan's 5 million. Compare that with Flagsmith, where every client SDK initialisation and every server-rendered page view can be a billable request. A team of 12 that would pay $720 a month on Flagsmith Scale-Up billed monthly pays $120 on ConfigCat Pro billed monthly.
The trade-off is ownership and depth. ConfigCat is a proprietary hosted service, and the Dedicated plan, at $4,500 a month billed annually, is the route to a private cloud or on-premise install. Experiments mean percentage splits whose results you read in Amplitude, Mixpanel or Google Analytics through its integrations.
- Unlimited seats, MAU and flag reads on every plan
- One price per plan, not per request
- Integrations with Amplitude, Mixpanel, Datadog, Jira and Terraform
- ✓The bill does not move with traffic or headcount
- ✓A short swap for flags plus remote config
- ✓50% startup programme for eligible early-stage teams
- ✗Proprietary service; private install from $4,500 a month, billed annually
- ✗Free plan limited to 10 flags and 2 environments
- ✗A/B results are read in your analytics tool
| Criterion | ConfigCat | Flagsmith |
|---|---|---|
| Billing unit | Flat plan, config downloads | API requests and team members |
| Seats | Unlimited on every plan | 1, 3, then 5 to 20 |
| Entry paid plan, monthly | $120/mo | $45/mo |
| Licence | Proprietary service | BSD-3-Clause |
Pick ConfigCat when you want a fixed bill and unlimited seats for flags and remote config. If you need to run the software yourself without a $4,500 contract, or real A/B statistics, GrowthBook is the better move.
LaunchDarkly
LaunchDarkly is the hosted incumbent of feature management. Its release controls run deep: approval requests on Foundation, then custom roles, unlimited audit history, required approvals, scheduling, code references and bulk targeting on Enterprise. Experiments, multi-armed bandits and observability are part of the same platform.
Its bill is built from two units. A service connection is one instance of one server-side SDK connected to one environment, measured in minutes: 5 are free, then $10 each per month billed yearly ($12 monthly). A client-side MAU is a unique user seen by a browser or mobile SDK in a month: 1,000 are free, then $8.33 per 1,000 billed yearly ($10 monthly). LaunchDarkly's own example, 10 connections and 2,000 MAU, comes to $58.33 a month billed yearly. Seats are unlimited, even on the free plan.
The limit is that the bill follows your architecture. Ten microservices, each with three replicas in two environments, already make 60 service connections, so $550 a month billed yearly before a single user is counted. Pre-production environments count too, and Ruby or Python servers that fork worker processes start one SDK instance per process. The Relay Proxy, a small Go application you run in your own infrastructure, cuts outbound connections, not billed ones. The platform itself is delivered as LaunchDarkly's hosted service, so check this first if you moved from Flagsmith Cloud to an on-premise install. Zapier users can also wire it into other tools through our LaunchDarkly Zapier integration guide.
- Approval requests from the Foundation plan
- Unlimited seats on every plan
- Experiments, bandits and observability built in
- ✓Approvals, custom roles and code references
- ✓Transparent per-unit rates for Foundation
- ✓Documented OpenFeature providers and a Relay Proxy
- ✗Bill grows with services, replicas and environments
- ✗Runs as a hosted service, Relay Proxy aside
- ✗Custom roles and unlimited audit history need Enterprise
| Criterion | LaunchDarkly | Flagsmith |
|---|---|---|
| Billing unit | Service connections and client-side MAU | API requests and team members |
| Seats | Unlimited | 1, 3, then 5 to 20 |
| Where it runs | Hosted, Relay Proxy in your infra | Cloud, private cloud or self-hosted |
| Free plan projects | 1 project, 3 environments | 1 project, unlimited environments |
Pick LaunchDarkly when governance and auditability outweigh price and everything can run in the vendor's cloud. If you run dozens of services or must self-host, Unleash gives approvals and SSO without counting connections.
Statsig
Statsig is a hosted experimentation platform with feature flags, dynamic configs, product analytics and session replay. Its pricing is the opposite of Flagsmith's: flag and config checks are unlimited on the Developer and Pro plans, and seats are unlimited too. What you pay for are metered events, the data points that feed metrics and experiments. The Developer plan includes 2 million a month for free; Pro is $150 a month with 5 million included, then $0.05 per 1,000.
One detail changes the maths. An exposure becomes a metered event when your app checks a flag that has Metric Lifts enabled, the setting that measures a flag's impact. Keep Metric Lifts off on plain release toggles and kill switches, and turn it on only where you want the measurement, or a free flag becomes a billed one. Fully shipped or disabled flags generate no metered events. Two smaller rules move the count: exposures are deduplicated per user and flag over 60 minutes on client SDKs but only 1 minute on server SDKs, and every event dimension beyond the first is billed as one more event.
Two limits keep it lower in our ranking. Change reviews and approvals start on Pro, while SSO, role-based access control and teams need Enterprise, so governance costs more than on Unleash. And ownership has moved twice: OpenAI agreed to buy Statsig in 2025, then in 2026 Amplitude acquired the product, brand and customer base while the original team stayed at OpenAI. If long-term ownership matters to your procurement, ask Amplitude about the roadmap. It runs as a hosted service, with a warehouse-native deployment on Enterprise.
- Unlimited flag and config checks on every plan
- Advanced statistics: CUPED, sequential tests, switchback tests
- Unlimited seats
- ✓Flags never billed per check
- ✓Holdouts, bandits and CUPED in the product
- ✓Generous free plan with 2M events a month
- ✗SSO and RBAC only on Enterprise
- ✗Metric Lifts exposures count as billed events
- ✗Hosted service; owned by Amplitude since 2026
| Criterion | Statsig | Flagsmith |
|---|---|---|
| Billing unit | Metered events | API requests and team members |
| Flag checks | Unlimited | Counted as requests |
| Entry paid plan | $150/mo | $45/mo monthly |
| Where it runs | Hosted, warehouse native on Enterprise | Cloud, private cloud or self-hosted |
Pick Statsig when experiments are the job and you want flag checks off the bill. If you need SSO or role-based access without an Enterprise contract, or an install on your own servers, choose Unleash or GrowthBook.
DevCycle
DevCycle is a hosted feature flag platform built around OpenFeature, the open standard for flag evaluation, and it supports OpenFeature across all of its SDKs. Dynatrace bought the DevCycle platform, and since 2026 a Dynatrace subsidiary provides the service. Every plan has unlimited seats, flags, environments and projects, and A/B testing is included from the free tier.
Billing counts client-side MAU, config requests and events. A MAU is a unique user ID with at least one client-side SDK initialisation in a month. The Free plan covers 1,000 MAU, 10,000 cloud config requests and 100,000 server config requests a month. Business is $500 a month billed annually ($625 monthly) for 100,000 MAU, 1,000,000 cloud and 10,000,000 server config requests, then $2.50 per extra 1,000 MAU. Audit logging, roles and permissions, and a custom domain proxy come with Business.
The jump is the weak point. Between 1,000 free MAU and the $500 Business plan there is no middle tier, where Flagsmith steps from $45 to $300 billed monthly. Approval workflows, full RBAC and SAML SSO need Enterprise. It runs as a hosted service, and the move to Dynatrace is recent, worth checking before a multi-year commitment. On Business, events past 500,000 a month cost $0.5 per 10,000, though the tracking built into the SDKs is not billed.
- OpenFeature support across all SDKs
- Unlimited seats, flags and projects on every plan
- Flag importer and Terraform provider
- ✓Moving off it later is a provider swap
- ✓No seat fees for product and QA
- ✓A/B testing on the free plan
- ✗Nothing between Free and $500 a month
- ✗Approval workflows and SSO need Enterprise
- ✗Hosted service, billed by active users
| Criterion | DevCycle | Flagsmith |
|---|---|---|
| Billing unit | Client-side MAU and config requests | API requests and team members |
| Seats | Unlimited | 1, 3, then 5 to 20 |
| Mid-tier price, monthly | $625/mo (Business) | $300/mo (Scale-Up) |
| Where it runs | Hosted service | Cloud, private cloud or self-hosted |
Pick DevCycle when OpenFeature is a requirement and your active users fit its MAU tiers. Below 100,000 MAU, ConfigCat Pro at $110 to $120 a month costs a fraction of the Business plan.
7 Flagsmith alternatives compared
The seven alternatives side by side, with the billing unit that decides each bill. Flagsmith's reference points: Start-Up at $45 and Scale-Up at $300 a month billed monthly.
| Best for | What you pay for | Free plan | Team size | Visit | ||||
|---|---|---|---|---|---|---|---|---|
| 1 | PostHog | Best if you left for analytics | Flag requests, 1M free a month | 4.5/5 | Free, then from $0.0001/request | ✓ | Product teams of any size | Visit → |
| 2 | GrowthBook | Best open source with experiments | Seats on cloud; free self-hosted | 4.4/5 | $40/seat/mo (Pro) | ✓ | Data-driven teams up to 30 users on Pro | Visit → |
| 3 | Unleash | Best for flags at scale with governance | Seats; MAU and connections unlimited | 4.3/5 | $75/seat/mo | ✓ | Engineering orgs, 5 seats and up | Visit → |
| 4 | ConfigCat | Best fixed price, unlimited seats | Flat plan by config downloads | 4.1/5 | $110/mo yearly | ✓ | Small and mid-size teams | Visit → |
| 5 | LaunchDarkly | Best for enterprise governance | Service connections and client-side MAU | 4.1/5 | $10/connection past 5 (yearly) | ✓ | Large engineering organisations | Visit → |
| 6 | Statsig | Best free plan for experiments | Metered events; flag checks free | 3.9/5 | $150/mo (Pro) | ✓ | Startups running many tests | Visit → |
| 7 | DevCycle | Best OpenFeature-native option | Client-side MAU and config requests | 3.8/5 | $500/mo yearly (Business) | ✓ | Teams standardising on OpenFeature | Visit → |
Prices are the entry paid plan as each vendor lists it in 2026, in US dollars. Yearly prices are per month, billed for twelve months up front.
Is your tool missing from this page? Email us: hello@hackceleration.com
You leave Flagsmith because… take this
1M free requests a month, same SDK for flags and events.
You self-host and want real A/B statisticsGrowthBookMIT core, unlimited users self-hosted; CUPED on Pro or Enterprise.
Your request bill grows faster than your teamUnleash$75 a seat, MAU and connections unlimited.
You need approvals and SSO without Scale-Up pricingUnleashChange requests and SSO on Pay-As-You-Go.
You want one fixed price and the whole company on itConfigCat$110 a month yearly, unlimited seats.
Compliance wants the deepest audit and approval controlsLaunchDarklyRequired approvals and unlimited audit history on Enterprise.
You run dozens of tests on a startup budgetStatsigFlag checks never billed, 2M free events.
OpenFeature is now a requirementDevCycleOpenFeature support across all SDKs.
Should you stay with Flagsmith?
Keep Flagsmith if…
- You self-host the open source build and only need flags and remote config: it has no API limits, where Unleash OSS stops at 1 project and 2 environments.
- Your traffic sits under 1,000,000 requests a month and your team is 3 people: Start-Up at $45 a month billed monthly costs less than Unleash ($375 floor self-hosted), ConfigCat Pro ($120 monthly) or Statsig Pro ($150).
- You need an on-premise or air-gapped install with vendor support: Flagsmith Enterprise covers cloud, private cloud and self-hosted, up to air-gapped instances.
- You rely on its OpenFeature providers and its MCP server, and moving would mean rebuilding segments that the export does not carry.
Switch if…
- Your app runs past 1,000,000 requests a month and pays the Start-Up overage of $7 per 100,000: Unleash includes 53 million on Cloud and ConfigCat counts cached downloads, not checks.
- You need change requests or SSO and do not want to jump from $45 to $300 a month: Unleash Pay-As-You-Go includes both.
- More than 20 people need access: Scale-Up stops at 20 team members, while ConfigCat, LaunchDarkly, Statsig and DevCycle have unlimited seats.
- Your product team wants experiment results without building a pipeline: PostHog, GrowthBook and Statsig read the impact of each flag in the tool.
How to choose a Flagsmith alternative
A feature flag tool lets you change what your software does without deploying new code. You wrap a feature in a flag, ship the code with the flag off, then turn it on for 1% of users, a beta segment or one customer, and roll back in seconds if errors climb. Most tools also store remote config, values such as a price or a timeout that you change from a dashboard. Flagsmith does both, which is why the shortlist above, like our wider ranking of feature flag tools, mixes flag specialists (Unleash, ConfigCat, LaunchDarkly, DevCycle) and experimentation platforms with flags (PostHog, GrowthBook, Statsig).
How a flag is evaluated decides your bill. In remote evaluation, the SDK sends the user's attributes to the vendor and gets the result back: simple, but each call is a request. In local evaluation, a server SDK downloads all flag definitions, keeps them in memory and decides on its own, so one poll replaces thousands of calls. Client SDKs in a browser or a mobile app usually fetch the flags for one user at start-up, and an edge proxy or CDN can serve them close to the user. Ask every vendor which of these its SDKs support for your languages.
Then translate each billing unit into your own numbers. A request (Flagsmith, PostHog) is one call to the flag API. A MAU, monthly active user (LaunchDarkly, DevCycle), is one unique user seen by a client SDK in a month, so anonymous visitors count. A service connection (LaunchDarkly) is one server SDK instance connected to one environment, so replicas and environments multiply it. A seat (Unleash, GrowthBook) is one person who logs in to the dashboard. A config download (ConfigCat) is one fetch of the cached config file, however many checks follow. A metered event (Statsig) is a data point that feeds a metric.
Before you sign, demand four things: approval workflows on the plan you can afford, an audit log, SSO if your company has an identity provider, and a stale flag view, because flags left in code turn into debt. Write your flag checks against OpenFeature, so the next move is a provider swap. If your team also wants in-product analytics, our product analytics ranking compares the tools you would plug experiments into, and marketers testing landing pages rather than code will find their tools in our heatmap and A/B testing ranking.
Moving off Flagsmith without breaking production
Export each environment from the project settings, then rebuild segments and overrides by hand: the export does not carry them. Push partial rollouts to 0% or 100% before you switch, because each vendor buckets users with its own hashing, so a 30% rollout will not hit the same people elsewhere. Run both tools side by side behind one wrapper for a few days and log any answer that differs. Our LaunchDarkly migration guide details the shadow-mode method, which works in either direction.
Open source, open core or source available
BSD-3-Clause (Flagsmith) and MIT (GrowthBook's core, PostHog outside ee/) let you host and modify the code with few obligations. AGPL-3.0 (Unleash) asks you to publish changes you serve over a network. Flipt's server uses the Fair Core License with an MIT future clause, a different deal from OSI open source, and its Pro plan costs $200 a month billed monthly ($2,000 a year) for up to 10 instances. In Flagsmith, GrowthBook and Unleash alike, governance sits in the paid tiers.
Rollback guardrails and AI assistants
The first newer feature is the automated guardrail: GrowthBook's Safe Rollouts with auto-rollback on Pro, and Unleash's impact metrics, automated progression and safeguards on its paid plans, watch a rollout so a bad release stops without someone awake to click. The second is the MCP server, which lets an AI assistant read or change flags: Flagsmith, GrowthBook and DevCycle (from its free plan) ship one. Treat a flag change from an assistant as a production change and start with a read-only key. If you work in Claude, our PostHog connector for Claude page shows a connector that can create and update feature flags.
If you run Harness or want flags in Git
Harness FME (Feature Management and Experimentation, the former Split) makes sense if your pipelines already run on Harness, where it is one of the six modules of the DevOps plan. For flags kept in Git with pull-request reviews, Flipt syncs its UI with Git and, on Pro, opens pull requests from the UI, while GO Feature Flag runs from a YAML file and a Docker container, on top of OpenFeature. For the release side, see our CI/CD tools ranking and the developer stack we recommend around it.
- Count your monthly requests, client-side MAU and server instances per environment before comparing any price.
- Check which plan includes change requests, SSO and audit logs: it is often two tiers above the entry plan.
- Ask whether server SDKs in your languages support local evaluation.
- Read the licence file in the repository, not the marketing page, before calling a tool open source.
- List the segments and overrides you will rebuild after the Flagsmith export.
- Set a billing limit or usage alert on day one with usage-based vendors.
Flagsmith alternatives, the FAQ
Is Flagsmith still open source?
Yes, mostly. The bulk of its code is under BSD-3-Clause, a permissive licence, and a few repositories are MIT. Role-based access control, SAML/SSO and certain database integrations are closed source and need a licence, so the governance layer of a self-hosted Flagsmith is a paid add-on.Can I export my flags out of Flagsmith?
Yes, per environment, from the Export tab of the project settings, with an optional filter by tag. The file lists flags, states and multivariate values. Exports stay available for only two weeks, so download a local copy. On import, the Skip strategy keeps existing flags and Overwrite Destructive replaces them, segment and identity overrides included.Who owns Flagsmith, and where is its data stored?
Flagsmith is published by Bullet Train Ltd. Its terms place user data processing and storage in the UK, with application data replicated across 8 AWS regions (US, EU, UK, Australia, Brazil, India, Singapore, Korea). A co-founder wrote in 2025 that most of its revenue comes from on-premise deployments.How do I cancel a Flagsmith subscription?
Give notice before the end of the current period. A subscription renews automatically for the same duration unless either side gives notice, no fees are refunded for a period you do not renew, and access stops as soon as the subscription ends. Flagsmith then has no obligation to keep your data, so export your environments before that date.What is OpenFeature, and does it make switching easier?
Yes, for the code. OpenFeature is an Apache 2 licensed open standard, a CNCF incubating project, that puts a vendor-neutral API between your code and your flag tool. You call OpenFeature in your code and plug in a provider for Flagsmith, Unleash, LaunchDarkly or GrowthBook. Changing vendor then means changing the provider, while segments and rules still need rebuilding.Is Unleash's open source edition still free to self-host?
Yes, the server stays free under AGPL-3.0, its licence since version 8.0.0, while the official Docker images are still Apache 2.0. What changed is Unleash Edge: its open source version is deprecated in favour of the Enterprise edition, with long-term support since the end of 2025 and end of life at the end of 2026. Self-hosters who relied on open source Edge for client-side traffic need a new plan.What happened to Split?
Harness bought it in 2024, and it is now sold as Harness Feature Management and Experimentation (FME). It counts usage in Monthly Tracked Keys, the unique keys evaluated for treatments in a month, plus impressions and events, and Harness says FME is moving away from seat-based pricing towards user-based billing.How do I stop feature flags from becoming technical debt?
Give each flag an owner and a removal date when you create it, then review stale flags every sprint. Several tools do the spotting for you: Unleash has a dedicated stale flag dashboard on its paid plans, GrowthBook has stale flag management and archiving, and DevCycle adds stale flag detection on Business.Do I need a paid feature flag service at all?
Not at first: a handful of release toggles fit in a config value. The free middle ground is wider than most teams think. Flagsmith's self-hosted open source build has no API limits, Unleash Open Source has no seat limit and holds 5,000 flags per instance, and Flagsmith's cloud Free plan includes up to 100,000 users in experiments. Pay when you need approvals, SSO or a support SLA.