Featurevisor

Featurevisor keeps feature flags, experiment variations and remote configuration as files in a Git repository. Flag changes then travel through the same branch, review and history that application code does, which removes the separate dashboard where toggles get flipped with no trail behind them.

The flow has three steps. A project is defined declaratively in the repository, a CI pipeline builds static JSON datafiles and uploads them to a CDN or a server of your choosing, and SDKs inside the application fetch a datafile and evaluate features against the context the application supplies. A project is assembled from attributes as the inputs for conditions, segments as reusable targeting rules, features carrying flags, variations and variables, and global variables with their own targeting and lifecycle.

The use cases the project documents follow from that structure.

• Progressive delivery: roll a feature out gradually instead of shipping it to everyone at once.
• Decoupled releases: separate deploying the code from releasing the feature it carries.
• Experiments: run A/B and multivariate tests through variations.
• Entitlements: gate features by role, and test in production safely.
• Remote configuration: change typed values without a deploy.
• Ownership: establish who owns each feature, with dependency management and safe deprecation.

SDKs cover JavaScript, React, React Native, Vue, Next.js, Go, Python, PHP, Ruby, Java, Kotlin, Swift, Rust and Elixir, with OpenFeature providers for most of them. An official agent skill teaches tools such as Claude Code and Cursor how to create features, write rollout rules and debug evaluations, and a local catalog command reviews every change in the browser.