WebMCP: a new standard for AI to interact with websites through structured tools¶
Insight: WebMCP is a proposed standard enabling AI systems to interact with websites through structured, machine-readable tool interfaces rather than traditional bot-style clicking or full DOM parsing. The hosts debate imperative vs. declarative API design: whether developers should provide step-by-step instructions for AI agents or declare capabilities that agents discover and execute autonomously. Key advantage is token efficiency — reduced computational overhead compared to sending full DOM representations.
Detail: WebMCP appears to be the web-specific implementation of broader MCP principles, allowing developers to expose their apps as structured tools for AI consumption. The standard aims to replace the current approach of AI navigating web interfaces visually (slow, brittle, token-expensive) with defined machine-readable endpoints. This relates to the broader MCP UI discussion (Kent C. Dodds on Syntax #973) about when to return HTML from tool calls vs. keeping responses as data.
Sources
Related: mcp-ui-web-next-form in mcp.md — COMPLEMENTS