Annotations
Annotations are optional metadata fields that can be added to MCP tools, resources, and templates. They help provide hints to the AI model about how to handle specific data.
Common Annotations
audience: Specifies who the tool is intended for (e.g.,userorassistant).priority: Indicates the importance of a tool or resource relative to others.labels: Arbitrary tags that can be used for grouping or filtering by the AI client.
Annotatons enhance the "semantics" of the protocol, allowing for more intelligent tool selection.
Questions & Answers
What are annotations in the context of the Model Context Protocol?
Annotations are optional metadata fields that can be included with MCP entities like tools, resources, and templates to provide hints to AI models on how to best utilize or handle them.
How can the `audience` annotation be used?
The audience annotation specifies who the tool or resource is intended for, such as user for direct interaction or assistant for the model's internal use.
How do annotations improve tool selection by AI clients?
By adding semantic weight and descriptive metadata (like priority and labels), annotations allow AI clients to more intelligently group, filter, and prioritize which tools to present to the model or user.