GrowthBook is an open-source platform for feature flags, A/B experimentation, and product analytics that product, data, and platform teams can either consume through a hosted cloud account or run entirely on their own infrastructure.

What it is

GrowthBook is a TypeScript codebase that combines feature flagging, experimentation, and product analytics in one application. It ships with a hosted option at app.growthbook.io and a self-hosted deployment that runs from a Docker Compose file in the repository, serving the UI on port 3000. The project describes itself as an open core product: the bulk of the code sits under the permissive MIT license, while several directories fall under the separate GrowthBook Enterprise License. Around it sits an ecosystem of client libraries, documentation, warehouse connectors, and an API surface for automation.

The problem it targets is stated directly in its own philosophy section: the top one percent of companies spend thousands of hours building experimentation, feature flagging, and analytics platforms in-house, while the rest either pay for expensive third-party SaaS tools or assemble unmaintained open-source libraries. GrowthBook replaces that SaaS bill and that patchwork with a single deployable platform whose experimentation statistics and metric definitions are computed against the customer's own data warehouse. It lives in the product experimentation and continuous delivery toolchain, alongside the data stack rather than beside it.

Key capabilities

• Feature flags with advanced targeting, gradual rollouts, and experiments attached to the same flag.
• 24 SDKs, including React, Python, Android (Kotlin), and iOS (Swift).
• Experiment statistics engine implementing CUPED, Sequential testing, Bayesian analysis, Post-Stratification, Bandits, and SRM checks.
• Warehouse-native querying of 11 data sources, including BigQuery, Snowflake, and Databricks.
• Flexible SQL-backed metric definitions covering conversion rates, ratios, and quantiles.
• Built-in Product Analytics suite for dashboards, plus screenshots, custom meta fields, and Markdown documentation on entities.
• Webhooks, a full REST API, and an MCP server for creating features, starting experiments, and cleaning up stale flags.

Who uses it and how

• Teams that want experimentation on their own infrastructure self-host with docker compose up -d, pulling a prebuilt public Docker image that needs no registry credentials.
• Data organisations running warehouse-native setups point GrowthBook at BigQuery, Snowflake, Databricks, or ClickHouse and keep metric logic in SQL they control.
• Product engineering teams instrument React, Python, Kotlin, and Swift clients through the 24 published SDKs, then roll out changes gradually by flag.
• Platform and data engineering groups automate the lifecycle through the REST API, webhooks, and the MCP server instead of the web UI.
• Smaller teams start on a free GrowthBook Cloud account before deciding whether to move to self-hosting.

Getting started

Clone https://github.com/growthbook/growthbook.git, run docker compose up -d, and visit http://localhost:3000; building the image from the repository Dockerfile instead requires a one-time docker login dhi.io because its hardened base images come from Docker's dhi.io registry. A free GrowthBook Cloud account is the alternative for teams that do not want to operate the stack.

How it compares

This entry names no directly comparable tools, so GrowthBook stands alone in this registry. Its README frames the competitive choice in categories rather than products: an in-house platform the company builds itself, an expensive third-party SaaS tool, or a set of unmaintained open-source libraries.

When to use it — and when not to

Self-hosting means operating the full Docker Compose stack and, for warehouse-native experiments, maintaining a live connection and metric definitions in a data source such as BigQuery, Snowflake, or Databricks; teams without a warehouse or without Docker capacity should use the hosted cloud option or look elsewhere. The licence field reads NOASSERTION because the repository is open core, so anyone planning commercial use should read the LICENSE file carefully before assuming MIT terms apply throughout. The project also carries 793 open issues, which suggests a large and active backlog worth weighing against how quickly a given team needs fixes.

git clone https://github.com/growthbook/growthbook.git
cd growthbook
docker compose up -d