APIbenchmarks
PostHog logo

PostHog

PostHog · Ranked #3 of 7 in Feature Flag APIs

85.3/ 100
AExcellent

Open-source product OS where feature flags ride alongside analytics, session replay and experiments; flags are priced per evaluation request with a huge free allowance.

Best for

All-in-one product analytics with built-in flags

Screenshot of PostHog

Overview

PostHog Feature Flags is the flag-management and rollout module inside PostHog's open-source product OS, which bundles analytics, session replay, A/B testing, surveys, and a data warehouse. Unlike standalone vendors such as LaunchDarkly or Split, PostHog does not sell feature flags as a discrete product, they ship as one tab in the same tool that already holds your behavioral data. That tight coupling is the core selling point: a flag's exposure can be wired directly to funnels, retention, and experiments without exporting events to a separate analytics stack. The flag engine itself is mature, boolean and multivariate flags, percentage rollouts, property- and group-based targeting, JSON payloads (remote config), scheduled changes, flag dependencies, and early-access feature management, and it is genuinely usable for free up to 1M flag requests/month, which covers most early-stage teams entirely.

Architecturally, PostHog has invested heavily in flag performance and resilience. Server-side local evaluation lets SDKs download flag definitions into memory and resolve flags with zero network I/O, cutting evaluation latency from roughly 500ms (remote /flags call) to 10-50ms, and bootstrapping eliminates client-side load flicker. In 2025 the team rewrote the flag service from Python to Rust. The economics are also aggressive at scale: per-request pricing falls 10x from $0.0001 (1-2M tier) to $0.00001 above 50M requests, making PostHog one of the cheaper options for high-volume flag evaluation, especially since flags piggyback on an account you may already run for analytics.

The honest caveats: reliability has been bumpy. Between October 21-30, 2025 the flags service suffered four incidents totaling over 14 hours of major impact, three sharing a root cause of CPU under-sizing and a missing CPU-alerting gap, documented transparently in PostHog's own public post-mortem. The product also carries a real learning curve; reviewers repeatedly note the UI is dense and that documentation, while attractive, thins out for advanced flag scenarios. Teams wanting a focused, battle-hardened, enterprise-SLA flag vendor with deep approval-workflow governance may still prefer a dedicated player, but for product teams already invested in (or attracted to) the all-in-one PostHog suite, the flags are a strong, well-priced, open-source-backed option.

How this score is derived

The APIbenchmarks Index is a weighted sum of four dimensions, each scored on an absolute 0–100 reference scale. See the methodology for every mapping.

DimensionScoreWeightContribution
Documentation & DXDocs are visually polished and cover SDK setup and local evaluation well, but reviewers consistently say they thin out for advanced flag debugging, and the AI helper draws from the same shallow advanced material.
88
30%26.4
ReliabilityFeature Flags API shows 100% uptime on the public status page over the trailing 3 months, but a transparently-documented Oct 2025 cluster of four incidents (14+ hrs major impact) shows resilience is still maturing.
82
25%20.5
Ecosystem & SDKsFlags are embedded in a broad open-source product suite (analytics, replay, experiments, warehouse) with 18+ official SDKs and tight native integration into experiments and behavioral data.
80
25%20.0
AccessibilityGenerous free tier (first 1M flag requests/month free, no card), transparent usage-based pricing, and a self-hostable open-source codebase lower the barrier, though the dense UI raises the practical learning curve.
92
20%18.4
APIbenchmarks Index (ABI)85.3

Table 1. Derivation of the ABI for PostHog. Contribution = score × weight; the index is their sum.

At a glance

Vendor
PostHog
Pricing model
Per request (volume-tiered)
Free tier
First 1M flag-eval requests/mo free, no credit card
Official SDKs
12 languages

Pricing

Free$0/moFirst 1,000,000 feature-flag requests per month free; no credit card required; full flag feature set.
Pay-as-you-go (1-2M tier)$0.0001 / requestPer-request rate applied to flag-evaluation requests beyond the 1M free monthly allowance.
Pay-as-you-go (2-10M tier)$0.000045 / requestVolume-discounted rate for 2-10M monthly flag requests.
Pay-as-you-go (10-50M tier)$0.000025 / requestVolume-discounted rate for 10-50M monthly flag requests.
Pay-as-you-go (50M+ tier)$0.00001 / requestLowest tier, a 10x reduction vs the entry rate, for 50M+ monthly flag requests.

Key features

  • Boolean and multivariate feature flags
  • Percentage-based gradual rollouts
  • User and group (account-level) property targeting
  • Server-side local evaluation (in-memory, no network call)
  • Client bootstrapping to eliminate page-load flicker
  • Remote config / JSON payloads attached to flags
  • Scheduled flag changes for time-based rollouts
  • Feature flag dependencies
  • Early-access feature management and opt-in
  • Native A/B testing and experiment integration with statistical significance

Official SDKs

Web / JavaScriptNode.jsPythonRubyGoRustPHPJava.NETiOSAndroidReact

Strengths & trade-offs

Strengths
  • +Generous free tier, first 1M flag-evaluation requests per month are free with full feature access and no credit card
  • +Server-side local evaluation cuts flag latency from ~500ms to 10-50ms by removing network I/O
  • +Flags are natively wired into PostHog analytics, experiments, and session replay, no event export needed to measure flag impact
  • +Aggressive volume pricing: per-request cost drops 10x from $0.0001 to $0.00001 above 50M requests
  • +Open-source and self-hostable, with 18+ official SDKs covering web, mobile, and major backend languages
  • +Rich flag model: multivariate flags, JSON payloads/remote config, group targeting, scheduled changes, and flag dependencies
Trade-offs
  • Reliability has wobbled, four flag-service incidents in late Oct 2025 totaling 14+ hours of major impact (CPU under-sizing root cause)
  • Dense, technical UI with a real learning curve for designing flags, cohorts, and experiment variants
  • Documentation is strong for getting started but thin for advanced flag debugging scenarios
  • Local evaluation can cause performance issues in edge/lambda and stateless PHP environments without an external cache
  • Not a dedicated flag vendor, teams wanting deep governance/approval workflows and enterprise SLAs may find it less specialized
  • Request-based billing means high-traffic client-side evaluation can get expensive without adopting local evaluation

What developers say

G2 4.6/5 (feature management category)

Users praise the all-in-one integration of flags with analytics and the flexible targeting, while criticism centers on a steep learning curve, dense UI, and reliability hiccups rather than the flag feature itself.

PostHog feels like a genuinely all-in-one product: analytics, session replay, and feature flag management all in one place, without pushing you into a bunch of separate tools.

Key figures

Feature Flags API uptime (trailing ~3 months)100.00%PostHog public status page
Flag evaluation latency with local evaluation~10-50ms (down from ~500ms remote)PostHog engineering blog
Free monthly flag requests1,000,000 requests/moPostHog pricing page
Entry per-request price (1-2M tier)$0.0001 / requestPostHog pricing page
High-volume per-request price (50M+ tier)$0.00001 / requestPostHog pricing page
Oct 2025 flag-service major-impact downtime14+ hours across 4 incidentsPostHog public post-mortem

Compare PostHog head to head

Sources

  1. https://posthog.com/feature-flags
  2. https://posthog.com/docs/feature-flags
  3. https://posthog.com/pricing
  4. https://posthog.com/docs/feature-flags/local-evaluation
  5. https://posthog.com/blog/how-we-improved-feature-flags-resiliency
  6. https://posthog.com/handbook/company/post-mortems/2025-10-21-feature-flags-recurring-outages
  7. https://www.posthogstatus.com/us
  8. https://www.g2.com/products/posthog/reviews
  9. https://softwarefinder.com/analytics-software/posthog/reviews

Figures last verified 2026-06-27. Spotted an error? corrections@apibenchmarks.com