GrowthBook MCP server

What the model can do

GrowthBook feature flags and experiments, on GrowthBook Cloud or your own server: flags with their environments, rules and change history, experiments with status, phases and results, projects, environments, metrics and segments. Can turn a flag on or off per environment, create, edit, archive, revert, publish a draft of or delete a flag, and create, edit, start, stop, archive, refresh results of or comment on an experiment.

Reads

• growthbook_list_featuresGrowthBook feature flags: id, description, project, value type, default value, owner, tags…
• growthbook_get_featureOne GrowthBook feature flag with its rules: each rule's type (force, rollout, experiment), environments, targeting condition, value…
• growthbook_feature_historyThe change history of GrowthBook feature flags: revisions with version, status (published, draft, pending review…), comment…
• growthbook_list_experimentsGrowthBook experiments, most recently updated first: name, tracking key, status (draft, running, stopped), project, owner, hypothesis, variations…
• growthbook_experiment_resultsThe latest analysed results of one GrowthBook experiment: users per variation and, per metric, each variation's mean, percent change against control…
• growthbook_list_projectsGrowthBook projects: id, name, description and when each was last updated.
• growthbook_list_environmentsGrowthBook environments (production, staging…): id, description, whether production…
• growthbook_list_metricsGrowthBook metrics used to judge experiments: id, name, type (binomial, count, duration, revenue), whether higher is better, description, tags…
• growthbook_list_segmentsGrowthBook segments (groups of users experiments can be analysed by): id, name, description, identifier type, kind and projects.

Acts, when you ask

These change something in GrowthBook. The model is told to use them only when you ask for exactly that, and each one is granted by name: per team on a shared connection, by the owner on a personal one.

• growthbook_add_experiment_commentAdds a comment to a GrowthBook experiment's discussion thread, credited to the connection's own identity (the token's user…
• growthbook_toggle_featureTurn a GrowthBook feature flag on or off in the named environments and publish the change at once…
• growthbook_create_featureCreate a GrowthBook feature flag, off in every environment (Memtro lists each environment as disabled so no "default state" switches it on).
• growthbook_update_featureEdit a GrowthBook flag's description, owner, default value, tags or project, publishing a new revision at once.
• growthbook_set_feature_archivedArchive a GrowthBook flag (it leaves the SDK payload, so applications fall back to their own default) or unarchive it.
• growthbook_revert_featureRestore a GrowthBook flag to an earlier published revision (version numbers from growthbook_feature_history) and publish the result at once as a new…
• growthbook_publish_feature_revisionPublish a draft revision of a GrowthBook flag (version from growthbook_feature_history with status draft), making its changes live at once.
• growthbook_delete_featurePermanently delete a GrowthBook feature flag and all of its revisions.
• growthbook_create_experimentCreate a GrowthBook experiment as a draft (nothing is served until it is started).
• growthbook_update_experimentEdit a GrowthBook experiment's name, hypothesis, description, owner, tags or metrics.
• growthbook_start_experimentStart a GrowthBook experiment, so its variations are served to live users (or stage it, if it has a scheduled start).
• growthbook_stop_experimentStop a running GrowthBook experiment and record its result (won, lost, inconclusive, or dnf for did not finish).
• growthbook_set_experiment_archivedArchive a GrowthBook experiment (it leaves the SDK payload) or unarchive it.
• growthbook_refresh_experiment_resultsRun a fresh results analysis for a GrowthBook experiment now.

Connect it

First connect GrowthBook on Dashboard → Connections, then add Memtro to your client once. The GrowthBook tools appear alongside the tools of everything else you have connected; nothing is configured per connector.

claude plugin marketplace add memtro-org/claude-memtro
claude plugin install memtro@memtro --config api_key=sk-memtro-...

{
  "mcpServers": {
    "memtro": {
      "url": "https://platform.memtro.com/api/mcp",
      "headers": {
        "Authorization": "Bearer sk-memtro-..."
      }
    }
  }
}

{
  "mcp": {
    "memtro": {
      "type": "remote",
      "url": "https://platform.memtro.com/api/mcp",
      "headers": {
        "Authorization": "Bearer sk-memtro-..."
      }
    }
  }
}

Who may use it

The key identifies the person, and the server offers that person the tools they may use everywhere else in Memtro. A shared GrowthBook connection is used by the teams an administrator granted it to, tool by tool, on the Permissions page; a personal connection is used by its owner alone, and its tools that act stay off until the owner allows them. See Teams and permissions and Connectors.

Ask it like this

• › Is new-checkout on in production?
• › How is the pricing experiment doing?
• › Who changed the checkout flag last week?
• › Which experiments are running in the web project?
• › Stop the pricing experiment and record variation B as the winner.

Questions

Is this an official GrowthBook MCP server?

No. It is Memtro's MCP server with Memtro's GrowthBook connector behind it: the connector reads GrowthBook through its API with the credentials you connect on Dashboard → Connections, and GrowthBook publishes nothing here. The same endpoint carries every other system you connect and Memtro's memory tools.

What does it cost?

Nothing on top of Memtro. One price per organisation by its number of users, from $50 a month for up to 10 users after a 7-day free trial, with every connector, the MCP server, jobs and memory included. You bring your own model keys.

Can it change anything in GrowthBook?

Only when you ask, and only what it has been granted: add a comment to an experiment's discussion; turn a feature flag on or off in named environments, publishing at once; create a feature flag, off in every environment; edit a flag's description, owner, default value, tags or project; archive or unarchive a feature flag; revert a flag to an earlier revision, or publish a draft revision; delete a feature flag and all its revisions permanently; create a draft experiment, or edit its name, hypothesis, description, owner, tags or metrics; start an experiment, or stop it and record the result; archive or unarchive an experiment; run a fresh results analysis for an experiment (queries the data warehouse). Every other tool reads, searches or lists. An administrator grants each acting tool by name per team, and a personal connection acts only after its owner allows it.

Which clients work?

Any MCP client that speaks Streamable HTTP: Claude Code, Cursor, Claude Desktop, OpenCode, Codex, Zed, Windsurf, Cline and the Anthropic and OpenAI SDKs. There is no local process to run; the server is remote and authenticated with your key.

Other MCP servers · Engineering and operations

• GitHub MCP server
• AWS CloudWatch MCP server
• Sentry MCP server
• ServiceNow MCP server
• Okta MCP server
• Twilio MCP server