RELATED 01
dbt MCP Server
OFFICIAL SERVERdbt Labs' server for running dbt commands, querying metrics and reading project metadata
617GitHub stars
76k/moPyPI downloads
81popularity score
Optimizely Experimentation MCP Server
Hosted Optimizely server for querying and managing experiments, feature flags, audiences and results
Official · optimizely.com Remote (hosted) · Streamable HTTP Data and analyticsDeveloper tools
Connect your client
Remote (hosted)
Add in one click
Opens the app, which asks before it adds the server.
Or add it by command or config
Claude CodeTerminal
claude mcp add --transport http optimizely-experimentation https://exp.mcp.opal.optimizely.com/mcpClaude (desktop and web)Settings → Connectors → Add custom connector
https://exp.mcp.opal.optimizely.com/mcpChatGPTSettings → Apps → Advanced → Developer mode, then Create app
https://exp.mcp.opal.optimizely.com/mcpCursor~/.cursor/mcp.json
{
"mcpServers": {
"optimizely-experimentation": {
"url": "https://exp.mcp.opal.optimizely.com/mcp"
}
}
}VS Code.vscode/mcp.json
{
"servers": {
"optimizely-experimentation": {
"type": "http",
"url": "https://exp.mcp.opal.optimizely.com/mcp"
}
}
}Codex~/.codex/config.toml
[mcp_servers.optimizely-experimentation]
url = "https://exp.mcp.opal.optimizely.com/mcp"Gemini CLITerminal
gemini mcp add --scope user --transport http optimizely-experimentation https://exp.mcp.opal.optimizely.com/mcpWhat it does
This is Optimizely's own hosted MCP server for its Experimentation products (Feature Experimentation and Web Experimentation). There is nothing to install: a client connects to https://exp.mcp.opal.optimizely.com/mcp over streamable HTTP and the user signs in with their Opti ID through OAuth, picking their Opal instance during sign-in. The server then sees the same projects, flags, experiments and data that the account can see in the Optimizely UI.
Its eight tools fall into three groups. Query tools return entity schemas, run structured queries over projects, flags, experiments and environments, summarise one experiment's results, and pull program-level reporting (lift, significance, confidence intervals, conversion rates, visitor counts) for paused, concluded and archived experiments. One manage tool creates, updates and archives flags, experiments, audiences and other entities, with a companion tool that returns the templates and required fields for those changes. An implement tool searches the Feature Experimentation SDK docs across languages such as JavaScript, Python, Java, Go, Swift, React and Next.js.
Optimizely notes that changes made through the manage tool hit live Optimizely data. Its install page also states that using this server does not consume Opal credits, a point the quickstart leaves out.
What it needs
An Opti ID login and an Optimizely account with Opal enabled, connected to at least one Feature Experimentation or Web Experimentation instance. An MCP client that supports remote (streamable HTTP) servers with OAuth, such as Claude Code, Claude Desktop, Claude.ai, Cursor, VS Code with Copilot, Windsurf or Codex. The host exp.mcp.opal.optimizely.com must be reachable through any corporate proxy or VPN.
Limits
Hosted only: there is no package, image or source to run yourself. For Web Experimentation, variation-level code changes (custom HTML, CSS and JavaScript) cannot be made through it; Optimizely points to the Visual Editor instead. exp_program_reporting_top_experiments reads the reporting tool, so it is most useful for paused and concluded experiments rather than running ones. Sign-in sessions are time-limited and the client asks for OAuth again when one expires. Access is bounded by the signed-in account's existing Optimizely permissions.
Who it suits
Suits teams already on Optimizely Feature or Web Experimentation with Opal enabled who want to ask about experiments, flags and results, or get SDK setup help, from an editor or chat client without building API scripts. Because it is hosted and signs in with OAuth, there are no tokens to manage, but it is no use without an Optimizely account and Opal, and it cannot run offline or on your own infrastructure. Since one tool can change live flags, experiments and audiences, users who only want reporting should review each change the client proposes before approving it.
Tools 8
exp_get_schemas- Returns the data schema of experimentation entities such as projects, flags and experiments.
exp_execute_query- Runs a structured query against the account's experimentation data.
exp_summarize_test_result- Fetches and analyses the results of a single experiment.
exp_program_reporting_top_experiments- Lists the best-performing experiments across the program from reporting data on paused and concluded tests.
exp_program_reporting_execute_query- Runs custom queries over results of paused, concluded and archived experiments, such as one metric across many tests.
exp_search_fx_sdk_docs- Searches Feature Experimentation SDK documentation for the supported languages and frameworks.
exp_manage_entity_lifecycle- Creates, updates or archives flags, experiments, audiences and other entities.
exp_get_entity_templates- Returns templates and required fields for creating or updating entities.