Comparison
Railflag vs Split (Harness FME)
Split was bought by Harness in 2024 and is now Harness Feature Management & Experimentation (FME), sold as a module of the Harness platform. It is a mature product with strong experimentation. Railflag is a smaller, focused service: a published price, approvals on every plan, and data in the EU. Here is how to choose.
Choose Harness FME if you use Harness for CI/CD, need its experimentation statistics, or need mobile SDKs. Choose Railflag if you want a published price without a sales process, approvals on every plan, rollouts paused automatically and all data in the EU.
At a glance
IncludedPaid tiers only, add-on or limitedNot available
| Railflag | Harness FME (Split) | |
|---|---|---|
| Free plan | Starter: 3 members, 3 environments, 10,000 browser users a month, never cut off | Free plan inside Harness; limits not published |
| Published prices | Team €249 and Business €899 a month; browser users above the plan at a published rate | Quoted, by seat and monthly tracked keys |
| No per-seat fees | Unlimited members on paid plans | Priced per seat |
| Evaluates flags in your process | Always, from rules held in memory; keeps serving if it can't reach us | Server SDKs, from rules held in memory |
| Changes reach SDKs in real time | Streamed within a second, polling as a fallback | Streamed, with polling available |
| OpenFeature providers | Node, browser, Python, Go and Java | Eleven providers, including Java, Node, Python, Go, .NET, iOS and Android |
| Approval before a change goes live | Every plan, the free one too; Production requires one by default | Approvers per environment; plans not published |
| Audit log | Every plan, the free one too | Per flag, environment and admin; retention not published |
| Pauses a rollout when metrics go bad | Release Watch on Team and Business reads Datadog, Prometheus, New Relic or Sentry | Alerts on guardrail metrics; the kill is manual unless wired to a Harness pipeline |
| MCP server for AI agents | Hosted; agents can request Production changes, never approve them | Harness MCP server with feature management tools |
| Data stored in the EU | Always: stored in the EU (Amsterdam) | No EU residency option published for FME |
| Self-hosting | Hosted only | Hosted; only proxy components run on your side |
| Built-in experiment statistics | Consistent assignment; analysis in your own analytics tool | Statistics engine with sequential testing, included |
| Mobile SDKs | Not yet | iOS, Android, React Native and Flutter |
Checked against Harness FME (Split)'s public pricing and documentation on 19 November 2026 (see their pricing). Vendors change quickly. If a row above is out of date, tell us and we will correct it.
When Split is the better choice
- You run on Harness already. FME connects to Harness pipelines and policy-as-code, so flags become part of the same delivery process.
- Experiments matter as much as releases. Harness FME includes a statistics engine with sequential testing and warehouse-native experiments.
- You need SDKs Railflag doesn't have, such as iOS, Android, React Native and Flutter.
When Railflag is the better choice
- You want a price without a sales call. Railflag publishes its plans: free Starter, Team €249 and Business €899 a month, with unlimited members. Harness FME is quoted, by seat and tracked keys.
- You don't want the rest of a platform. Railflag is a flag service you can adopt in an afternoon, not a module of a larger suite.
- You want bad rollouts paused for you. Release Watch pauses a rollout itself when Datadog, Prometheus, New Relic or Sentry show drift. Harness FME alerts on guardrail metrics; stopping is manual unless you wire a pipeline.
- Data must stay in the EU. Railflag stores everything in Amsterdam. Harness publishes no EU residency option for FME.
Moving from Split
Railflag's automated migration supports AB Tasty and Unleash today, not Split. A move from Split is done by hand or with our engineers, and the order matters:
- Put flag checks behind OpenFeature while still on Split, using one of Harness FME's OpenFeature providers. After that, switching is a provider change. See the OpenFeature guide.
- Clean up first. Remove feature flags that have been fully on or off for weeks.
- Recreate flags, treatments and segments in Railflag. Split treatments map to Railflag variations, and attributes to the evaluation context.
- Treat percentage rollouts and experiments separately. Every platform hashes users differently, so finish in-progress experiments on Harness.
- Switch one service and one environment at a time, Staging first.
Want help? Assisted migration is a one-off €4,900, credited in full against your first year of Business; choose "Something else" as your current tool.
Frequently asked questions
- Is Split still sold on its own?
- No. Split is now Harness Feature Management & Experimentation, and split.io's pricing page redirects to Harness. FME is enabled inside a Harness account.
- Can Railflag import my flags from Split automatically?
- Not yet. Railflag's automated migration supports AB Tasty and Unleash today. Moving from Split is a manual or assisted process; our engineers can plan it with you.
- Do I have to change the code that checks flags?
- If your code calls flags through OpenFeature, only the provider changes. If it calls the Split SDK directly, moving to OpenFeature first lets you switch later without touching call sites again.
Try Railflag next to Split
Start free, point one service's OpenFeature provider at Railflag in Staging, and compare.
Start for free