diff --git a/pkg/agent/agent_media.go b/pkg/agent/agent_media.go
index a773d2ebb..866f7dc24 100644
--- a/pkg/agent/agent_media.go
+++ b/pkg/agent/agent_media.go
@@ -75,6 +75,14 @@ func resolveMediaRefs(messages []providers.Message, store media.MediaStore, maxS
 				continue
 			}
 
+			if strings.HasPrefix(mime, "audio/") {
+				dataURL := encodeImageToDataURL(localPath, mime, info, maxSize)
+				if dataURL != "" {
+					result[i].Audio = append(result[i].Audio, dataURL)
+				}
+				continue
+			}
+
 			pathTags = append(pathTags, buildPathTag(mime, localPath))
 		}
 
diff --git a/pkg/providers/protocoltypes/types.go b/pkg/providers/protocoltypes/types.go
index f3553f8b0..07aa086e0 100644
--- a/pkg/providers/protocoltypes/types.go
+++ b/pkg/providers/protocoltypes/types.go
@@ -76,6 +76,7 @@ type Message struct {
 	Content          string         `json:"content"`
 	Media            []string       `json:"media,omitempty"`
 	Attachments      []Attachment   `json:"attachments,omitempty"`
+	Audio            []string       `json:"audio,omitempty"`
 	ReasoningContent string         `json:"reasoning_content,omitempty"`
 	SystemParts      []ContentBlock `json:"system_parts,omitempty"` // structured system blocks for cache-aware adapters
 	ToolCalls        []ToolCall     `json:"tool_calls,omitempty"`
@@ -91,4 +92,4 @@ type ToolFunctionDefinition struct {
 	Name        string         `json:"name"`
 	Description string         `json:"description"`
 	Parameters  map[string]any `json:"parameters"`
-}
+}
\ No newline at end of file