claude-design-mcp

The materials explicitly state that no keys or environment variables are required, and there is no request for API tokens, account credentials, or other sensitive authentication data, so credential exposure and misuse risk appears low.

No remote endpoints are declared, and the materials do not state that user data is sent to external services. Based on the available facts, there is no clear data egress path; however, due to missing documentation, actual network behavior should still be verified after installation.

The system checks explicitly mark this tool as executes-code, indicating it can run code locally or drive local processes. This is a common MCP capability and not by itself a high-risk red flag, but its runtime environment and callable privileges should be constrained.

The description says it can create, generate, iterate, and export design systems, so it is reasonable to infer it may read or write local project or design output files. The current materials do not define the exact access scope, and there is no clear evidence of over-privilege, but directory access should still be granted under least-privilege principles.

The project is open source and MIT-licensed, which are clear risk-reducing factors; however, it comes from a third-party registry, has 0 stars, unknown maintenance status, and no README, leaving limited evidence of maturity and auditability, so it should be used only after dependency pinning and source review.