Optimizely MCP. Control feature flags, start or pause experiments.

Optimizely MCP Server: A/B Testing and Feature Flag Management.

Forget logging into your dashboard just to check if an experiment is running or toggle a flag. This server connects your AI client straight into Optimizely, letting you audit your entire digital property stack using nothing but natural conversation. You can manage everything—from listing every single project you own to pausing live tests and checking audience definitions—without ever touching the platform UI.

System Overview and Project Mapping

You need a clear map of what's running? This server gives it to you right out of the gate. It retrieves a full list of all Optimizely projects available in your account using list_projects. Need details on one specific workspace? Use get_project to pull all metadata for any given project.

Experiment Management and Control

Managing tests shouldn't be a pain. You can get an immediate overview of multiple active or draft experiments across a whole project using list_experiments. For deep dives, use get_experiment to grab every detail on one specific test—you'll see its configuration, its current status, and how it’s set up.

When you need action, this server lets you control the rollout instantly. You can immediately pause an experiment using pause_experiment, stopping whatever active user traffic was hitting those tests. Conversely, if you want to restart a test that got stuck or needs a fresh run, just use start_experiment. These tools let you manage live tests and rollouts via chat.

Feature Flag Management

Feature flags are key, but tracking them can be a mess. You can pull a list of every feature flag name available in your current project using list_feature_flags and then grab the full definition and status for any single flag with get_feature_flag. This lets you audit whether those flags are set up correctly before code hits production.

Audience and Event Auditing (The Setup Check)

Before you deploy anything, you gotta make sure your tracking is solid. The server lists every defined user group or audience within a project using list_audiences. This lets you verify exactly who's being targeted for any given test. You can also list all configured conversion tracking events for the entire project via list_events, confirming that your setup can track what it needs to.

This server gives your agent direct access to audit every core resource—projects, feature flags, audiences, and conversion events. It lets you check experiment status or toggle a flag's operational state using natural talk. You don’t need tokens or complex API calls; you just ask your AI client what you want done.