Experiments and feature flags are now on Free and Growth
Ever since we launched experiments and feature flags last year for our Enterprise customers, I’ve received the same question over and over: That’s great—when are they coming to the other plans?
Well, today I’m excited to announce that experiments and feature flags are now available on Mixpanel's Free and Growth plans. They come with:
- Limited amounts of experiments and feature flags included
- Full feature parity with our Enterprise plan
- The ability to add even more experiments and feature flags simply by swiping your credit card
Why is Mixpanel doing experiments and feature flags?
AI has made it dramatically easier to produce code. Turning an idea into a working feature now takes an afternoon instead of a sprint. So building isn't the bottleneck; validation is. You still have to know, before you commit, if an idea actually holds up to the scrutiny of real users.
As we wrote in our guide The New Testing Paradigm: “The teams that thrive today aren’t the ones shipping the most. They’re the ones who are shipping the right things at a faster pace.”
That's what AI-native experiments and feature flags help you achieve, and we want every team to be able to use them in the same place they’re already measuring impact.
We can now release a feature, hypothesize its impact, and measure results immediately. Being able to do that without having to leave the tool that already has all of our data, all of our core analytics, and all of our reports just made sense.”
Run the whole loop via your harness of choice
Experiments and feature flags are fully supported in Mixpanel MCP and Mixpanel Agent, so the loop runs in the tools you're already working in:
- Identify opportunities. Ask what's happening in your onboarding funnel and get a root cause analysis against live data.
- Spin up the test. Your AI client proposes a hypothesis and a primary metric, then launches it.
- Ship the winner. An AI coding agent creates the flag, wires it in, and ramps the rollout.
And if you want something more specific than that, Mixpanel Headless gives you the primitives to build your own workflow.
Make confident calls
Testing only matters if you can trust the results. That’s why we’ve put a ton of investment into our statistical engine, and we include in every plan all the methodologies and health checks you need to ensure the lift you’re seeing is valid—no gating what you need just because you’re experimenting less.
- Choose between Frequentist, Sequential, and Bayesian (in beta) models
- Deploy CUPED, Benjamini-Hochberg correction, Winsorization, Bonferroni correction, SRM monitoring, and pre-experiment bias testing as needed
- Set success metrics and guardrail metrics, and customize your statistical settings as needed (e.g., changing confidence threshold)
Does this all sound like gobbledygook? No worries, Mixpanel Agent is there to help you properly set up your test and interpret the results.
Target precisely, ship safely
You can also run tests, create kill switches, and run rollouts right from Mixpanel with our built-in feature flags. Jump right from a rollout to the relevant Session Replay, or spin up a dashboard to track post-release success automatically.
We’ve built out feature flags to be as flexible, easy-to-use, and reliable as possible:
- SDKs and OpenFeature support. Flags are available across our client and server SDKs, with both remote and local evaluation depending on how latency-sensitive your use case is. Prefer a vendor-agnostic setup? Mixpanel also works with OpenFeature providers, so you can evaluate flags through OpenFeature's standard API and swap providers without rewriting your flag logic.
- Flag persistence. Client SDKs can persist assignments to the device with a configurable TTL, so users get a consistent experience even when the network drops, instead of falling back mid-session.
- Permissions, audit trail, and QA testers. Control who can create and edit flags at the project or per-flag level, see the full history of who changed what and when, and whitelist specific users to test a variant before it rolls out to anyone else.
What's included
Free: 1,000 Monthly Experiment Users (MEU)*, up to 10 active feature flags.
Growth: 5,000 MEU per month, up to 50 active feature flags. Scale to 100,000 MEU and 500 active flags by adding MEUs via Plan Builder.
*MEU counts users exposed to an experiment.
Get started for free now, and check out the Experiments docs and Feature Flags docs for full details.