Optimizely MCP

The Optimizely Experimentation MCP server gives AI assistants direct access to a feature flagging and experimentation project, so defining a test can happen next to the code that implements it. An agent can list projects, environments, flags, variations, rulesets and audiences, create or update a flag and its targeting rules, define experiment variations and traffic allocation, and read results including how each variation performed against a metric. Pairing it with a coding agent is the main use case. A developer adding a flag in the codebase can have the assistant create the matching flag and rollout rule in Optimizely, reference the right SDK key and environment, and later ask whether an experiment has reached a decision instead of switching to the results dashboard to read the numbers. Reading audience and event definitions also helps an agent write instrumentation that matches what the experiment expects to receive. It is meant for product engineers, growth and experimentation teams, and product managers running tests on Optimizely, especially those working inside an AI coding client where the flag and the code change belong to the same task. Authentication uses an Optimizely API token scoped to a project, so what an agent can read or modify follows the account's existing permissions rather than a blanket key.