readOnlyHint
A safety annotation in the WebMCP protocol that indicates to an AI agent that a tool has no side effects (it does not modify the UI, backend data or the physical world), thereby avoiding unnecessary user confirmation dialogs.
Timeline (last 90 days)
The Chrome team demonstrated three security annotations (hints) for WebMCP tool definitions: readOnlyHint, consequentialHint, and untrustedContentHint.
The readOnlyHint: true annotation tells the agent that the tool has no side effects (it does not change the UI, backend data, or the physical world).
MCP tools declare their own semantics through annotations (such as readOnlyHint and idempotentHint), and the client decides the retry strategy accordingly; only tools explicitly marked as read-only or idempotent are suitable for automatic retry when a connection is interrupted.
readOnlyHint and destructiveHint are metadata fields built into the MCP specification for communicating the degree of side effects of a tool to agent frameworks.
All Facts (4)
The Chrome team demonstrated three security annotations (hints) for WebMCP tool definitions: readOnlyHint, consequentialHint, and untrustedContentHint.
50%UnverifiedThe readOnlyHint: true annotation tells the agent that the tool has no side effects (it does not change the UI, backend data, or the physical world).
50%UnverifiedMCP tools declare their own semantics through annotations (such as readOnlyHint and idempotentHint), and the client decides the retry strategy accordingly; only tools explicitly marked as read-only or idempotent are suitable for automatic retry when a connection is interrupted.
50%UnverifiedreadOnlyHint and destructiveHint are metadata fields built into the MCP specification for communicating the degree of side effects of a tool to agent frameworks.
50%