What does AI recommend for feature flag platforms (agencies)? Checked 2026-10-22

Claude (Anthropic) with web search claude-opus-5 asked 2026-10-22

Coverage: Claude (Anthropic) with web search. Other engines not included.

The question asked, verbatim: best feature flag platforms for agencies. This page records one answer on one date. It is a measurement, not a ranking and not a recommendation.

Brands named, in the order the engine gave

1. per-seat mentioned
   Per-seat pricing and per-MAU pricing are what kill you.
2. ConfigCat recommended
   ConfigCat - my default recommendation for agencies.
3. Unleash mentioned
   Unleash - best pick when the client will self-host or has data-residency/EU requirements.
4. Statsig compared
   One comparison puts it as the cheapest entry point among full-featured platforms at ~$75/mo versus ~$110 for ConfigCat and ~$150 for Statsig.
5. Flagsmith mentioned
   Flagsmith - the handoff-friendly one.
6. PostHog mentioned
   PostHog - pick this when you're also delivering analytics or session replay, which agencies very often are.
7. GrowthBook mentioned
   GrowthBook - when the engagement is really experimentation (CRO, marketing-site A/B tests) rather than release safety.
8. LaunchDarkly recommended
   LaunchDarkly - not my first choice for agencies, but name it for enterprise clients.
9. OpenFeature mentioned
   Build against OpenFeature (the CNCF vendor-neutral flag SDK spec) rather than any vendor's SDK directly.
10. Amplitude mentioned
    - Amplitude: best feature flag tools (https://www.amplitude.com/compare/best-feature-flag-tools.md)

Sources the engine cited

The answer, as recorded

## The short answer For agency work the ranking is different than for a product team, because your constraints are: **many isolated client projects, rotating staff/contractors, and an eventual handoff where the client takes over the bill.** Per-seat pricing and per-MAU pricing are what kill you. So: **1. ConfigCat** - my default recommendation for agencies. Pricing is based on config JSON downloads, *not* per seat, per MAU, or per service connection, and team size is unlimited on every plan including free. That means you can put every dev, PM, and the client's own stakeholders in the dashboard at no extra cost - which is exactly the agency shape. It's organized into "products," so each client gets a clean boundary. Forever Free covers 10 flags / 2 environments / 2 products / 5M downloads, with Pro at $110/mo. ([pricing](https://configcat.com/pricing/), [why not per-seat](https://configcat.com/blog/2020/02/06/fair-pricing/)) **2. Unleash** - best pick when the client will self-host or has data-residency/EU requirements. Open-source core, and the cloud Enterprise tier includes unlimited projects. One comparison puts it as the cheapest entry point among full-featured platforms at ~$75/mo versus ~$110 for ConfigCat and ~$150 for Statsig. Project-level separation plus proper RBAC is the thing that makes multi-client tenancy actually safe. ([plans](https://www.getunleash.io/upgrade-unleash), [comparison](https://tech-insider.org/?p=18557)) **3. Flagsmith** - the handoff-friendly one. The self-hosted edition is free under BSD-3-Clause with no published API request ceiling, and *all* plans include unlimited flags and unlimited environments. You can stand it up inside a client's own infra, hand them the keys, and walk away with zero ongoing license obligation on either side. Cloud starts free (50k requests, 1 seat) and Start-Up is $45/mo. ([plans](https://www.costbench.com/software/feature-flags/flagsmith/), [top-7 writeup](https://flagsmith.com/blog/top-7-feature-flag-tools)) **4. PostHog** - pick this when you're also delivering analytics or session replay, which agencies very often are. 1M feature flag requests/month free, no per-seat charges, and flags sit next to the funnels you're already reporting to the client. Overage is cheap and tiered ($0.0001/request down to $0.00001), and you can set hard billing limits so a client's traffic spike doesn't become your invoice problem. ([flags pricing](https://posthog.com/feature-flags/pricing), [cutting costs](https://posthog.com/docs/feature-flags/cutting-costs)) **5. GrowthBook** - when the engagement is really *experimentation* (CRO, marketing-site A/B tests) rather than release safety. Open-source self-hosted is free and unlimited; Pro is $40/seat/mo. It has the real stats layer - CUPED, sequential testing, multi-armed bandits - which the pure flag tools don't, and it reads results off the client's existing warehouse. ([overview](https://www.amplitude.com/compare/best-feature-flag-tools.md)) **6. LaunchDarkly** - not my first choice for agencies, but name it for enterprise clients. It's the established reference platform combining flags, rollouts, experimentation, and observability, and large clients often already have a contract you can just get provisioned into. Avoid putting it on your own account: it's priced per MAU, which is unpredictable to pass through to a client. ([pricing](https://launchdarkly.com/pricing/), [MAU critique](https://statsig.com/perspectives/understanding-launchdarkly-pricing-is-it-worth-it)) ## The one thing I'd actually insist on Build against **OpenFeature** (the CNCF vendor-neutral flag SDK spec) rather than any vendor's SDK directly. ConfigCat, Flagsmith, Unleash, GrowthBook, and LaunchDarkly all have OpenFeature providers. For an agency this is the whole game: you write one integration pattern, reuse it across every client, and when a client insists on the platform their CTO already bought, you swap the provider instead of rewriting the codebase. It also means the handoff doesn't strand them on your tooling choice. Practical split if you want to stop reading here: **ConfigCat** as the house standard for client work, **Flagsmith or Unleash self-hosted** for regulated or on-prem clients, **PostHog** when analytics is in scope, and go where the client's existing contract is for enterprise. Sources: - [ConfigCat pricing](https://configcat.com/pricing/) · [ConfigCat on per-seat pricing](https://configcat.com/blog/2020/02/06/fair-pricing/) - [Unleash plans](https://www.getunleash.io/upgrade-unleash) · [LaunchDarkly vs Unleash vs GrowthBook](https://tech-insider.org/?p=18557) - [Flagsmith plans](https://www.costbench.com/software/feature-flags/flagsmith/) · [Top 7 feature flag tools](https://flagsmith.com/blog/top-7-feature-flag-tools) - [PostHog feature flags pricing](https://posthog.com/feature-flags/pricing) · [PostHog cutting flag costs](https://posthog.com/docs/feature-flags/cutting-costs) - [Amplitude: best feature flag tools](https://www.amplitude.com/compare/best-feature-flag-tools.md) - [LaunchDarkly pricing](https://launchdarkly.com/pricing/) · [Statsig on LaunchDarkly pricing](https://statsig.com/perspectives/understanding-launchdarkly-pricing-is-it-worth-it) - [Feature flag pricing models compared](https://rollgate.io/blog/feature-flags-pricing-comparison)

Recorded 2026-10-22 through Claude (Anthropic) with web search, model claude-opus-5. Permanent address: /answers/s/feature-flag-platforms-agencies-2026-10-22

Sources cited across the 21 recorded answers to this question

Alphabetical. A domain with a dated record links to it; the record says how often and when. Is one of these your site? Its record lists every recorded answer that cited it and carries the alert for the next citation.

amplitude.com · buildmvpfast.com · cadence.withremote.ai · capterra.com · configcat.com · costbench.com · dev.to · devtoollab.com · dupple.com · featureflip.io · flagshark.com · flagsmith.com · g2.com · getmonetizely.com · getunleash.io · github.com · growthbook.io · guptadeepak.com · harness.io · home.featbit.co · kameleoon.com · launchdarkly.com · openalternative.co · openfeature.dev · plushcap.com · posthog.com · rollgate.io · schematichq.com · statsig.com · stribog.com · tech-insider.org · techplained.com · trustradius.com · us.fitgap.com · userorbit.com · userpilot.com

Earlier recordings of this question

• asked 2026-10-20
• asked 2026-10-17
• asked 2026-10-15
• asked 2026-10-13
• asked 2026-10-12
• asked 2026-10-10
• asked 2026-10-07
• asked 2026-10-05
• asked 2026-10-04
• asked 2026-10-02
• asked 2026-09-30
• asked 2026-09-26
• asked 2026-09-24
• asked 2026-09-22
• asked 2026-09-18
• asked 2026-09-15
• asked 2026-09-13
• asked 2026-09-11
• asked 2026-09-09
• asked 2026-09-07

Markdown edition · all recorded answers · weekly source index · AEO Watch

Answer Watch records what one AI answer engine said on one date, nothing more. Each page names the engine and the model, carries the date the question was asked, and lists the brands that appeared in the answer in the order the engine gave them, with the sources it cited. It is a measurement, not a ranking, not a recommendation and not a claim about the world. Brands that did not appear in an answer are not named on any public page. Coverage: Claude (Anthropic) with web search. Other engines not included.