diff --git a/apps/tui-go/internal/app/server.go b/apps/tui-go/internal/app/server.go
index 26632bab..540f602a 100644
--- a/apps/tui-go/internal/app/server.go
+++ b/apps/tui-go/internal/app/server.go
@@ -92,6 +92,12 @@ func (m *Model) applyServer(msg protocol.ServerMessage) {
 	switch msg.Type {
 	case protocol.TypeSessionAnnounced:
 		m.onSessionAnnounced(msg)
+	case protocol.TypeRouterNarration:
+		entry := RouterEntry{Role: "narration_llm", Content: msg.Content}
+		if msg.LatencyMs != nil && msg.TotalTokens != nil {
+			entry.Metrics = &TurnMetrics{LatencyMs: *msg.LatencyMs, TotalTokens: *msg.TotalTokens}
+		}
+		m.appendRouter(msg.SessionID, entry)
 	case protocol.TypeChatTurn:
 		m.routerConnected = true
 		role := "router"
diff --git a/apps/tui-go/internal/app/view.go b/apps/tui-go/internal/app/view.go
index 5b2c8daf..19bc19e9 100644
--- a/apps/tui-go/internal/app/view.go
+++ b/apps/tui-go/internal/app/view.go
@@ -358,6 +358,19 @@ func (m Model) routerRows(w, h int) []string {
 			}
 		case "tool":
 			rows = append(rows, t.span("· tool output ("+itoaLen(e.Content)+" chars) — ^x to view", t.P.Dim))
+		case "narration_llm":
+			diamond := lipgloss.NewStyle().Foreground(t.P.Accent).Background(t.P.Bg).Bold(true).Render("◆")
+			lines := wrap(e.Content, w-2)
+			for i, ln := range lines {
+				if i == 0 {
+					rows = append(rows, diamond+" "+t.span(ln, t.P.FgStrong))
+				} else {
+					rows = append(rows, "  "+t.span(ln, t.P.FgStrong))
+				}
+			}
+			if s := metricsSuffix(e.Metrics); s != "" {
+				rows = append(rows, t.span(s, t.P.Faint))
+			}
 		case "narration", "stage":
 			rows = append(rows, t.span(e.Content, t.P.Dim))
 		}
diff --git a/apps/tui-go/internal/protocol/golden_test.go b/apps/tui-go/internal/protocol/golden_test.go
index df34c382..c6bb5568 100644
--- a/apps/tui-go/internal/protocol/golden_test.go
+++ b/apps/tui-go/internal/protocol/golden_test.go
@@ -99,6 +99,17 @@ func TestDecodeGoldenServerFrames(t *testing.T) {
 				}
 			},
 		},
+		{
+			name:      "router.narration carries content + metrics",
+			json:      `{"type":"router.narration","sessionId":"s1","content":"moving on","stageId":"validate","latencyMs":2000,"totalTokens":120,"sequence":10,"sessionSequence":5}`,
+			wantType:  TypeRouterNarration,
+			eventBear: true,
+			check: func(t *testing.T, m ServerMessage) {
+				if m.Content != "moving on" || m.LatencyMs == nil || *m.LatencyMs != 2000 || m.TotalTokens == nil || *m.TotalTokens != 120 {
+					t.Fatalf("router.narration: %+v", m)
+				}
+			},
+		},
 		{
 			name:      "snapshot_complete (non-event control frame)",
 			json:      `{"type":"snapshot_complete"}`,
diff --git a/apps/tui-go/internal/protocol/protocol.go b/apps/tui-go/internal/protocol/protocol.go
index 03d899b7..c5639fd1 100644
--- a/apps/tui-go/internal/protocol/protocol.go
+++ b/apps/tui-go/internal/protocol/protocol.go
@@ -44,6 +44,7 @@ const (
 	TypeModelChanged     = "model.changed"
 	TypeModelList        = "model.list"
 	TypeResourceStatus   = "resource.status"
+	TypeRouterNarration  = "router.narration"
 )
 
 // ServerMessage is a flat decode of every server->client variant. Field names
@@ -186,7 +187,8 @@ func (m ServerMessage) IsEventBearing() bool {
 		TypeInferenceStarted, TypeInferenceDone, TypeInferenceTimeout,
 		TypeToolStarted, TypeToolCompleted, TypeToolFailed, TypeToolRejected,
 		TypeToolAssessed,
-		TypeApprovalRequired, TypeApprovalResolved, TypeWorkspaceBound, TypeArtifactCreated:
+		TypeApprovalRequired, TypeApprovalResolved, TypeWorkspaceBound, TypeArtifactCreated,
+		TypeRouterNarration:
 		return true
 	default:
 		return false