Skip to content


What is flagd?

flagd is a feature flag evaluation engine. Think of it as a ready-made, open source, OpenFeature-compliant feature flag backend system.

With flagd, you can:

  • modify flags in real-time
  • define flags of various types (boolean, string, number, JSON)
  • use context-sensitive rules to target specific users or user traits
  • perform pseudorandom assignments for experimentation
  • perform progressive roll-outs of new features
  • aggregate flag definitions from multiple sources
  • expose aggregated flags as a gRPC stream to be used by in-process providers
  • expose OFREP service for configured flags

It doesn't include a UI, management console or a persistence layer. It's configurable entirely via a POSIX-style CLI. Thanks to its minimalism, it's extremely flexible; you can leverage flagd as a sidecar alongside your application, an engine running in your application process, or as a central service evaluating thousands of flags per second.

How do I deploy flagd?

flagd is designed to fit well into a variety of infrastructures and can run on various architectures. It runs as a separate process or directly in your application (see architecture). It's distributed as a binary, container image, and various libraries (see installation). If you're already leveraging containers in your infrastructure, you can extend the docker image with your required configuration. You can also run flagd as a service on a VM or a "bare-metal" host. If you'd prefer not to run an additional process at all, you can run the flagd evaluation engine directly in your application. No matter how you run flagd, you will need to supply it with feature flags. The flag definitions supplied to flagd are monitored for changes which will be immediately reflected in flagd's evaluations. Currently supported sources include files, HTTP endpoints, Kubernetes custom resources, and proto-compliant gRPC services (see syncs, sync configuration).

How do I use flagd?

flagd is fully OpenFeature compliant. To leverage it in your application you must use the OpenFeature SDK and flagd provider for your language. You can configure the provider to connect to a flagd instance you deployed earlier (evaluating flags over gRPC) or use the in-process evaluation engine to do flag evaluations directly in your application. Once you've configured the OpenFeature SDK, you can start evaluating the feature flags configured in your flagd definitions.