WebMCP lets a website describe selected actions as tools with clear names, descriptions, and parameters for compatible AI agents. This guide focuses on the Declarative API: start with an existing HTML form, test whether it can be called, and keep authentication and authorization in the backend.
Browsing: AI agent
An Agent Plugin is a directory format for packaging AI skills and MCP server configuration into a package that works across compatible agents. This article explains the structure of plugin.json, how to create a sample package, how to validate it, and which limitations to consider.
Small businesses can deploy AI agents safely by starting with three layers of control: granting the right permissions, requiring human approval for risky actions, and fully recording activity for review when needed.