WebMCP Glossary
Key terms and concepts for the AI-agent web and WebMCP standard.
WebMCP
Web Model Context Protocol — an emerging web standard that allows AI agents to discover and interact with website capabilities through structured tool definitions embedded in HTML.
Model Context Protocol (MCP)
A protocol that defines how AI models can access external tools, data sources, and services in a standardized way. WebMCP brings this concept to the browser.
Declarative API
A WebMCP approach where tools are registered using HTML form attributes like toolname and tooldescription, allowing AI agents to discover capabilities without executing JavaScript.
Imperative API
A WebMCP approach where tools are registered programmatically via navigator.modelContext.registerTool(), giving developers full control over tool definitions and behavior.
llms.txt
A plain-text file served at /llms.txt that provides AI models with a human-readable summary of a website's purpose, structure, and capabilities — similar to robots.txt but for LLMs.
AI Agent Readiness Score
A 0–100 score (graded A+ to F) that WebMCP Verify assigns to a website based on 15 checks across tool registration, AI discoverability, security, and best practices.
Tool Registration
The process of making a website's interactive capabilities discoverable by AI agents, either through declarative HTML forms or imperative JavaScript APIs.
Secure Context
A browser environment served over HTTPS, required for WebMCP APIs to function. Sites served over plain HTTP cannot use navigator.modelContext.
inputSchema
A JSON Schema object attached to an imperative tool registration that describes the expected input parameters, their types, and validation constraints.
Tool Annotations
Metadata hints like readOnlyHint and destructiveHint attached to tool registrations that help AI agents understand whether a tool modifies data or is safe to call speculatively.
Structured Data (Schema.org)
JSON-LD or microdata markup embedded in HTML that describes page content in a machine-readable format, improving discoverability by AI agents and search engines.
Content Security Policy (CSP)
An HTTP header that controls which scripts and resources a page can load. Overly restrictive CSP rules can block WebMCP tool registration scripts.