MCP (Model Context Protocol) is expanding rapidly across enterprises, creating a new wave of API sprawl similar to previous REST, GraphQL, gRPC, and Kafka expansions — but faster and wider. The challenge of 'seeing' and governing these APIs remains largely unsolved, and the common assumption that AI will handle discovery and documentation automatically is questioned. Machine-readable artifacts combined with LLMs like Claude, ChatGPT, and Gemini offer some promise for automated discovery and visualization, but new blind spots will emerge. The author is building tooling to generate side-by-side documentation for APIs, MCP servers, and Agent Skills, advocating for ephemeral and evolving approaches to API visibility rather than static ones.
Sort: