A/B testing and feature flags you run yourself.
Experimently is an open-source experimentation platform: experiments, feature flags, targeting rules and statistics, deployed on your own infrastructure so the assignment data never leaves it.
Apache-2.0. Both build profiles are the same licence and the same price: none.
Experiments and feature flags
Sticky assignment, gradual rollouts with configurable stages, mutual-exclusion groups and a global holdout, and automated safety rollback on error-rate and latency thresholds.
Targeting that is actually expressive
20+ operators — semantic version comparison, geo-distance, time windows, JSON path and array operations — compiled and cached, with throughput floors checked by a nightly benchmark job.
Statistics beyond a p-value
Frequentist and Bayesian analysis, sequential testing with mSPRT for early stopping, multi-armed bandits, and per-segment breakdowns with a Bonferroni correction across segments.
SDKs for the stack you have
JavaScript, React, React Native, Edge, Python, Go, Java, Android, iOS, Flutter, Ruby, PHP, .NET and Elixir, plus OpenFeature providers for JavaScript and Python, all MIT-licensed. The JavaScript SDK is on npm (npm install @getexperimently/js-sdk) and the Python SDK is on PyPI (pip install experimently), both as betas; the others are not on npm or the other package registries yet, and each of their guides shows how to install it from the repository. By default, flag evaluation and experiment assignment happen on the server. The JavaScript and Python SDKs can also evaluate flags locally (beta), tested against cases generated by the server's own evaluator.
Deploy it yourself
Run the published container images with Docker Compose on one machine, install them on Kubernetes with the Helm chart attached to each release, or use the CDK app to deploy to ECS Fargate, Aurora PostgreSQL and ElastiCache in your own AWS account. There is no hosted tier, so there is nothing to be locked into.
Open, and checkable
The whole platform is Apache-2.0 and the SDKs are MIT. The test suites, the automated checks and the release pipeline are all in the repository, so you can read them rather than take our word for it.
Run it locally
Self-host on KubernetesFour containers — Postgres, Redis, the API and the dashboard — with the database created and seeded on the way up.
git clone https://github.com/getexperimently/experimently.git
cd experimently
docker compose up -d --waitThen open localhost:3000. The quick start has the default credentials and what to change before anyone else can reach it.