feat: support audio content (#725)

Co-authored-by: Marcelo Trylesinski <marcelotryle@gmail.com>

src/mcp/server/fastmcp/prompts/base.py

@@ -7,9 +7,9 @@ from typing import Any, Literal
 import pydantic_core
 from pydantic import BaseModel, Field, TypeAdapter, validate_call
 
-from mcp.types import EmbeddedResource, ImageContent, TextContent
+from mcp.types import AudioContent, EmbeddedResource, ImageContent, TextContent
 
-CONTENT_TYPES = TextContent | ImageContent | EmbeddedResource
+CONTENT_TYPES = TextContent | ImageContent | AudioContent | EmbeddedResource
 
 
 class Message(BaseModel):

src/mcp/server/fastmcp/server.py

@@ -52,6 +52,7 @@ from mcp.server.streamable_http_manager import StreamableHTTPSessionManager
 from mcp.shared.context import LifespanContextT, RequestContext, RequestT
 from mcp.types import (
     AnyFunction,
+    AudioContent,
     EmbeddedResource,
     GetPromptResult,
     ImageContent,
@@ -275,7 +276,7 @@ class FastMCP:
 
     async def call_tool(
         self, name: str, arguments: dict[str, Any]
-    ) -> Sequence[TextContent | ImageContent | EmbeddedResource]:
+    ) -> Sequence[TextContent | ImageContent | AudioContent | EmbeddedResource]:
         """Call a tool by name with arguments."""
         context = self.get_context()
         result = await self._tool_manager.call_tool(name, arguments, context=context)
@@ -875,12 +876,12 @@ class FastMCP:
 
 def _convert_to_content(
     result: Any,
-) -> Sequence[TextContent | ImageContent | EmbeddedResource]:
+) -> Sequence[TextContent | ImageContent | AudioContent | EmbeddedResource]:
     """Convert a result to a sequence of content objects."""
     if result is None:
         return []
 
-    if isinstance(result, TextContent | ImageContent | EmbeddedResource):
+    if isinstance(result, TextContent | ImageContent | AudioContent | EmbeddedResource):
         return [result]
 
     if isinstance(result, Image):

src/mcp/server/lowlevel/server.py

@@ -405,7 +405,10 @@ class Server(Generic[LifespanResultT, RequestT]):
                 ...,
                 Awaitable[
                     Iterable[
-                        types.TextContent | types.ImageContent | types.EmbeddedResource
+                        types.TextContent
+                        | types.ImageContent
+                        | types.AudioContent
+                        | types.EmbeddedResource
                     ]
                 ],
             ],

src/mcp/types.py

@@ -657,11 +657,26 @@ class ImageContent(BaseModel):
     model_config = ConfigDict(extra="allow")
 
 
+class AudioContent(BaseModel):
+    """Audio content for a message."""
+
+    type: Literal["audio"]
+    data: str
+    """The base64-encoded audio data."""
+    mimeType: str
+    """
+    The MIME type of the audio. Different providers may support different
+    audio types.
+    """
+    annotations: Annotations | None = None
+    model_config = ConfigDict(extra="allow")
+
+
 class SamplingMessage(BaseModel):
     """Describes a message issued to or received from an LLM API."""
 
     role: Role
-    content: TextContent | ImageContent
+    content: TextContent | ImageContent | AudioContent
     model_config = ConfigDict(extra="allow")
 
 
@@ -683,7 +698,7 @@ class PromptMessage(BaseModel):
     """Describes a message returned as part of a prompt."""
 
     role: Role
-    content: TextContent | ImageContent | EmbeddedResource
+    content: TextContent | ImageContent | AudioContent | EmbeddedResource
     model_config = ConfigDict(extra="allow")
 
 
@@ -801,7 +816,7 @@ class CallToolRequest(Request[CallToolRequestParams, Literal["tools/call"]]):
 class CallToolResult(Result):
     """The server's response to a tool call."""
 
-    content: list[TextContent | ImageContent | EmbeddedResource]
+    content: list[TextContent | ImageContent | AudioContent | EmbeddedResource]
     isError: bool = False
 
 
@@ -965,7 +980,7 @@ class CreateMessageResult(Result):
     """The client's response to a sampling/create_message request from the server."""
 
     role: Role
-    content: TextContent | ImageContent
+    content: TextContent | ImageContent | AudioContent
     model: str
     """The name of the model that generated the message."""
     stopReason: StopReason | None = None