{"generated_at":"2026-10-14T20:07:22+00:00","item":{"added_at":"2026-08-12","analysis_method":"independent_manual_review_2026-10-10","assessment_status":"provisional","auth_posture":"auth_required","category":"Web Connector","confidence":"medium","description":"Manage feature flags, experiments, and product analytics in GrowthBook Cloud. Configure targeting, publish or retire flags, and launch, analyze, or stop experiments. OAuth access follows the authenticated user's organization permissions.","did":"d6420e17-1702-4e17-9197-18e67aab7f4f","documentation_url":"https://docs.growthbook.io/","editorial_description":{"extended_description":"Separate read and write tools inspect projects, environments, revisions, and dependencies or change production configuration. Built-in workflows provide guidance for approvals, revision conflicts, analysis, and permanent deletion; metric discovery supports analytics explorations.","schema_version":1,"summary":"Manage feature flags, experiments, and product analytics in GrowthBook Cloud.\nConfigure targeting, publish or retire flags, and launch, analyze, or stop experiments.\nOAuth access follows the authenticated user's organization permissions."},"homepage_url":"","icon":"","id":"d6420e17-1702-4e17-9197-18e67aab7f4f","installs":0,"last_scanned":"","license":"","long_description":"Separate read and write tools inspect projects, environments, revisions, and dependencies or change production configuration. Built-in workflows provide guidance for approvals, revision conflicts, analysis, and permanent deletion; metric discovery supports analytics explorations.","manual_assessment":"high","manual_review":{"all_findings":[{"evidence":"The Directory description says the connector can publish/ramp feature flags and launch/stop experiments through the write API and loaded skills.","reasoning":"Changing production feature targeting and experiments is High infrastructure control.","risk_type":"deployment_infrastructure","severity":"high","title":"Feature flags and experiments can be published or launched","triggering_tools_or_skills":["growthbook_api_write","growthbook_read_skill"]},{"evidence":"The description explicitly references safeguards for permanent deletion, but concrete delete actions are hidden behind skills/generic write API.","reasoning":"Permanent removal of production feature resources is destructive.","risk_type":"data_deletion","severity":"high","title":"Hidden workflows include permanent cleanup/deletion","triggering_tools_or_skills":["growthbook_api_write","growthbook_list_skills","growthbook_read_skill"]},{"evidence":"Only four top-level tools are listed while server-provided skills supply API contracts and workflows.","reasoning":"The full reachable write/delete/admin surface cannot be audited from this packet.","risk_type":"generic_gateway","severity":"high","title":"A generic write API and dynamically loaded skills hide concrete actions","triggering_tools_or_skills":["growthbook_api_write","growthbook_list_skills","growthbook_read_skill"]},{"evidence":"The connector injects mutable workflow instructions and organization data.","reasoning":"Those instructions/data should remain untrusted.","risk_type":"prompt_injection_exposure","severity":"low","title":"Server-provided skills and product data enter model context","triggering_tools_or_skills":["growthbook_read_skill","growthbook_api_read"]},{"evidence":"The connector runs at https://mcp.growthbook.io/mcp without a pinned server implementation in the packet.","reasoning":"The implementation can change independently of this review, so runtime verification and change monitoring remain necessary.","risk_type":"provenance_uncertainty","severity":"low","title":"GrowthBook MCP is a mutable hosted service","triggering_tools_or_skills":[]}],"assessment_status":"provisional","auth_and_scopes":"OAuth scopes actions to GrowthBook organizations and permissions of the authenticated user; exact API scopes and confirmation enforcement are not in the packet.","blocked_missing_evidence":null,"confidence":"medium","displayed_risk":"high","evidence_completeness":"partial","manual_overall_risk":"high","official_sources":[{"accessed_at":"2026-10-10T20:24:37Z","url_or_path":"plutosecurity/plutonium@b749c8dc43536c794409be9805edd48063f045e2:app/data/conn/claudesec/d6420e17-1702-4e17-9197-18e67aab7f4f.json","what_it_proves":"Candidate PR #125 record identity (name \"GrowthBook MCP\", publisher, type web_connector) and 4 stored tool names; no tool descriptions stored"},{"accessed_at":"2026-09-22T10:46:07Z","url_or_path":"https://api.anthropic.com/api/directory/servers?verified_tier=anthropic,partner,community (snapshot claude-directory-all-tiers-2026-09-22.json)","what_it_proves":"Official Claude Directory listing https://claude.ai/directory/d6420e17-1702-4e17-9197-18e67aab7f4f: complete tool_names inventory (4 tools), remote streamable-http endpoint https://mcp.growthbook.io/mcp, auth_posture auth_required, directory tier community"}],"primary_reason":"GrowthBook can change production feature flags/experiments and hides permanent deletion behind a generic API and dynamic skills.","provisional_risk":null,"publication_recommendation":"fix","requires_runtime_verification":true,"review_key":"claude:d6420e17-1702-4e17-9197-18e67aab7f4f","reviewed_at":"2026-10-10","reviewer_notes":"Reviewed all 4 top-level tools. Publication requires a pinned complete skill/action inventory, write schemas and confirmation matrix; do not present unspecified high-impact domains as safe.","risk_domains":{"arbitrary_network_access":"false","authz_ambiguity":"true","browser_control":"false","credential_or_admin_actions":"false","deletes_data":"true","deployment_or_infrastructure":"true","dynamic_hidden_actions":"true","filesystem_or_repo_access":"false","forwards_data_to_third_party":"true","modifies_data":"true","persistent_credentials":"true","prompt_injection_exposure":"true","provenance_uncertainty":"true","reads_private_data":"true","runs_code_or_queries":"false","sends_messages_as_user":"false","spends_or_moves_money":"false"},"tools_or_skills_reviewed":4,"tools_or_skills_total":4},"name":"GrowthBook MCP","needs_runtime_verification":true,"plutonium_url":"https://plutonium.pluto.security/detail.html?planet=claudesec&did=d6420e17-1702-4e17-9197-18e67aab7f4f&utm_source=plutonium_analysis_skill&utm_medium=claude_skill&utm_campaign=connector_risk_assessment","privacy_policy_url":"https://www.growthbook.io/legal/privacy-notice","provisional_risk":null,"publication_recommendation":"fix","publisher":"GrowthBook","publisher_url":"https://www.growthbook.io/","remote_url":"https://mcp.growthbook.io/mcp","repository_url":"","risk":"high","risk_severity":"high","security_risks":[{"description":"This connector can deploy code or change infrastructure and production environments. Review findings: Changing production feature targeting and experiments is High infrastructure control.","evidence":"Feature flags and experiments can be published or launched: The Directory description says the connector can publish/ramp feature flags and launch/stop experiments through the write API and loaded skills.","manual_risk_type":"deployment_infrastructure","remediation":{"block_tool_categories":["code_execution","state_change"],"steps":["Default to a sandbox or staging environment.","Require code review, human approval, audit logging, and a tested rollback path for production changes."]},"risk_type":"code_deployment","severity":"high","title":"Feature flags and experiments can be published or launched","triggering_tools":["growthbook_api_write","growthbook_read_skill"]},{"description":"This connector can permanently remove data or resources. Review findings: Permanent removal of production feature resources is destructive.","evidence":"Hidden workflows include permanent cleanup/deletion: The description explicitly references safeguards for permanent deletion, but concrete delete actions are hidden behind skills/generic write API.","manual_risk_type":"data_deletion","remediation":{"block_tool_categories":["destructive"],"steps":["Disable destructive tools unless deletion is an explicit workflow requirement.","Require human confirmation and verified backups before deletion."]},"risk_type":"deletes_data","severity":"high","title":"Hidden workflows include permanent cleanup/deletion","triggering_tools":["growthbook_api_write","growthbook_list_skills","growthbook_read_skill"]},{"description":"This connector can execute code, commands, queries, or dynamically selected actions. Review findings: The full reachable write/delete/admin surface cannot be audited from this packet.","evidence":"A generic write API and dynamically loaded skills hide concrete actions: Only four top-level tools are listed while server-provided skills supply API contracts and workflows.","manual_risk_type":"generic_gateway","remediation":{"block_tool_categories":["code_execution"],"steps":["Run only in a sandbox with least-privilege credentials and strict resource limits.","Disable generic execution gateways unless their reachable action set is fully understood.","Require approval before execution against production systems or sensitive data."]},"risk_type":"runs_code","severity":"high","title":"A generic write API and dynamically loaded skills hide concrete actions","triggering_tools":["growthbook_api_write","growthbook_list_skills","growthbook_read_skill"]}],"signature_status":"unknown","source_code_reviewed":false,"support_url":"support@growthbook.io","tags":["code_deployment","deletes_data","network_access","oauth_required","runs_code","web_connector"],"tools":[{"description":"","name":"growthbook_api_read"},{"description":"","name":"growthbook_api_write","risk":{"category":"code_execution","level":"high","recommendation":"Follow the connector-level remediation and require approval for consequential actions.","why":"This connector can execute code, commands, queries, or dynamically selected actions. Review findings: The full reachable write/delete/admin surface cannot be audited from this packet."}},{"description":"","name":"growthbook_list_skills","risk":{"category":"code_execution","level":"high","recommendation":"Follow the connector-level remediation and require approval for consequential actions.","why":"This connector can execute code, commands, queries, or dynamically selected actions. Review findings: The full reachable write/delete/admin surface cannot be audited from this packet."}},{"description":"","name":"growthbook_read_skill","risk":{"category":"code_execution","level":"high","recommendation":"Follow the connector-level remediation and require approval for consequential actions.","why":"This connector can execute code, commands, queries, or dynamically selected actions. Review findings: The full reachable write/delete/admin surface cannot be audited from this packet."}}],"tools_count":4,"transport":"streamable-http","type":"web_connector","url":"https://claude.ai/directory/d6420e17-1702-4e17-9197-18e67aab7f4f","uuid":"6b43296b-cf44-5fd5-b82a-a8d1ae6e6ab1","verification_tier":"community","version":""},"query_key":"growthbook mcp","schema_version":1,"status":"match"}