diff --git a/loop/loop.go b/loop/loop.go
index e115287a2b97683e179ceb6791a4303d1714228c..6d46338d55de9c0540d209370895f3eb985de418 100644
--- a/loop/loop.go
+++ b/loop/loop.go
@@ -245,6 +245,7 @@ func (l *Loop) processLLMRequest(ctx context.Context) error {
 	resp, err := llmService.Do(llmCtx, req)
 	if err != nil {
 		// Record the error as a message so it can be displayed in the UI
+		// EndOfTurn must be true so the agent working state is properly updated
 		errorMessage := llm.Message{
 			Role: llm.MessageRoleAssistant,
 			Content: []llm.Content{
@@ -253,6 +254,7 @@ func (l *Loop) processLLMRequest(ctx context.Context) error {
 					Text: fmt.Sprintf("LLM request failed: %v", err),
 				},
 			},
+			EndOfTurn: true,
 		}
 		if recordErr := l.recordMessage(ctx, errorMessage, llm.Usage{}); recordErr != nil {
 			l.logger.Error("failed to record error message", "error", recordErr)
diff --git a/loop/loop_test.go b/loop/loop_test.go
index c59f96f32160b451215b840ab5e20f3f6a2aec96..88d015447d6ac5c14fed23cc05efec4a9430c597 100644
--- a/loop/loop_test.go
+++ b/loop/loop_test.go
@@ -1548,6 +1548,11 @@ func TestProcessLLMRequestError(t *testing.T) {
 	if !strings.Contains(recordedMessages[0].Content[0].Text, "LLM request failed") {
 		t.Errorf("expected error message to contain 'LLM request failed', got: %s", recordedMessages[0].Content[0].Text)
 	}
+
+	// Verify EndOfTurn is set so the agent working state is properly updated
+	if !recordedMessages[0].EndOfTurn {
+		t.Error("expected error message to have EndOfTurn=true so agent working state is updated")
+	}
 }
 
 // errorLLMService is a test LLM service that always returns an error