diff --git a/provider/foundryprovider/agent.go b/provider/foundryprovider/agent.go
index 5407e076..a05bfeca 100644
--- a/provider/foundryprovider/agent.go
+++ b/provider/foundryprovider/agent.go
@@ -37,6 +37,10 @@ type AgentConfig struct {
 
 	// DisableStoreOutput disables service-side Responses output storage.
 	// Use this when local session history providers own conversation state.
+	// By default, store-disabled runs also request `reasoning.encrypted_content`
+	// so reasoning items can be replayed statelessly; use
+	// [openaiprovider.ResponsesIncludeReasoningEncryptedContent] in RunOptions to
+	// opt out.
 	DisableStoreOutput bool
 
 	// OpenAIOptions configure the OpenAI-compatible per-agent client. Foundry-owned
diff --git a/provider/foundryprovider/agent_test.go b/provider/foundryprovider/agent_test.go
index c15fa696..c9c55c30 100644
--- a/provider/foundryprovider/agent_test.go
+++ b/provider/foundryprovider/agent_test.go
@@ -10,6 +10,7 @@ import (
 
 	"github.com/microsoft/agent-framework-go/agent"
 	"github.com/microsoft/agent-framework-go/provider/foundryprovider"
+	"github.com/microsoft/agent-framework-go/provider/openaiprovider"
 	"github.com/openai/openai-go/v3/option"
 )
 
@@ -84,6 +85,33 @@ func TestNewAgentDisableStoreOutputSetsStoreFalse(t *testing.T) {
 	}
 }
 
+func TestNewAgentRunOptionCanDisableReasoningEncryptedContentInclude(t *testing.T) {
+	server := httptest.NewTLSServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		body := jsonMap(t, mustReadBody(t, r))
+		if body["store"] != false {
+			t.Fatalf("store = %#v, want false", body["store"])
+		}
+		if _, ok := body["include"]; ok {
+			t.Fatalf("include = %#v, want omitted", body["include"])
+		}
+		writeResponsesOK(w)
+	}))
+	defer server.Close()
+
+	foundryAgent := newFoundryAgent(t, server, foundryprovider.ModelDeployment("gpt-4o-mini"), foundryprovider.AgentConfig{
+		DisableStoreOutput: true,
+		Config: agent.Config{
+			RunOptions: []agent.Option{
+				openaiprovider.ResponsesIncludeReasoningEncryptedContent(false),
+			},
+		},
+	})
+
+	if _, err := foundryAgent.RunText(t.Context(), "hello").Collect(); err != nil {
+		t.Fatalf("RunText error = %v", err)
+	}
+}
+
 func TestNewAgentRunsAgainstFoundryAgentEndpoint(t *testing.T) {
 	server := httptest.NewTLSServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		if r.URL.Path != "/projects/proj/agents/my-agent/endpoint/protocols/openai/responses" {
diff --git a/provider/openaiprovider/chat.go b/provider/openaiprovider/chat.go
index 185a663f..037f8ccd 100644
--- a/provider/openaiprovider/chat.go
+++ b/provider/openaiprovider/chat.go
@@ -65,7 +65,10 @@ type AgentConfig struct {
 
 	// DisableStoreOutput is used only by [NewResponsesAgent]. It disables
 	// service-side output storage and prevents response IDs from being saved into
-	// agent sessions for later continuation.
+	// agent sessions for later continuation. While enabled, Responses requests
+	// also include `reasoning.encrypted_content` by default so reasoning items
+	// can be replayed statelessly; use [ResponsesIncludeReasoningEncryptedContent]
+	// to opt out per run.
 	// It is ignored by [NewChatCompletionsAgent].
 	DisableStoreOutput bool
 
diff --git a/provider/openaiprovider/responses.go b/provider/openaiprovider/responses.go
index 8470d1e8..38f98d71 100644
--- a/provider/openaiprovider/responses.go
+++ b/provider/openaiprovider/responses.go
@@ -72,11 +72,25 @@ func (o responsesNewParamsOpt) MAFValue() any {
 	return responses.ResponseNewParams(o)
 }
 
+type responsesIncludeReasoningEncryptedContentOpt bool
+
+func (o responsesIncludeReasoningEncryptedContentOpt) MAFValue() any {
+	return bool(o)
+}
+
 // ResponsesNewParams allows passing custom parameters to the underlying OpenAI Responses API calls.
 func ResponsesNewParams(params responses.ResponseNewParams) agent.Option {
 	return responsesNewParamsOpt(params)
 }
 
+// ResponsesIncludeReasoningEncryptedContent controls whether [NewResponsesAgent]
+// automatically requests `reasoning.encrypted_content` while Responses output
+// storage is disabled. The default is true so reasoning items can be replayed
+// across stateless turns.
+func ResponsesIncludeReasoningEncryptedContent(enabled bool) agent.Option {
+	return responsesIncludeReasoningEncryptedContentOpt(enabled)
+}
+
 func (a *responsesClient) formatOf(v any) (agent.ResponseFormat, error) {
 	return jsonformat.ForType(reflect.TypeOf(v))
 }
@@ -233,7 +247,8 @@ func responsesBuildCompletionParams(config AgentConfig, messages []*message.Mess
 		if param.IsOmitted(params.Store) {
 			params.Store = openai.Bool(false)
 		}
-		if !slices.Contains(params.Include, responses.ResponseIncludableReasoningEncryptedContent) {
+		if include, ok := agent.GetOption(opts, ResponsesIncludeReasoningEncryptedContent); (!ok || include) &&
+			!slices.Contains(params.Include, responses.ResponseIncludableReasoningEncryptedContent) {
 			params.Include = append(params.Include, responses.ResponseIncludableReasoningEncryptedContent)
 		}
 	}
diff --git a/provider/openaiprovider/responses_test.go b/provider/openaiprovider/responses_test.go
index 67be1ef0..66b18d4c 100644
--- a/provider/openaiprovider/responses_test.go
+++ b/provider/openaiprovider/responses_test.go
@@ -5979,6 +5979,55 @@ func TestResponsesNewParamsStoreFalseDoesNotDuplicateReasoningInclude(t *testing
 	}
 }
 
+func TestResponsesIncludeReasoningEncryptedContentFalseSkipsAutomaticInclude(t *testing.T) {
+	const input = `
+		{
+			"store":false,
+			"model":"gpt-4o-mini",
+			"input":[{
+				"type":"message",
+				"role":"user",
+				"content":[{"type":"input_text","text":"hello"}]
+			}]
+		}
+		`
+
+	const output = `
+		{
+			"id": "resp_67890",
+			"object": "response",
+			"created_at": 1741891428,
+			"status": "completed",
+			"model": "gpt-4o-mini-2024-07-18",
+			"output": [{
+				"type": "message",
+				"id": "msg_67d32764fcdc8191bcf2e444d4088804058a5e08c46a181d",
+				"status": "completed",
+				"role": "assistant",
+				"content": [{"type": "output_text", "text": "Hello!", "annotations": []}]
+			}]
+		}
+		`
+
+	server := newTestResponsesServer(t, input, output)
+	defer server.Close()
+
+	a := openaiprovider.NewResponsesAgent(
+		openai.NewClient(option.WithBaseURL(server.URL)),
+		openaiprovider.AgentConfig{
+			Model: "gpt-4o-mini",
+		},
+	)
+
+	_, err := a.RunText(t.Context(), "hello",
+		openaiprovider.ResponsesNewParams(responses.ResponseNewParams{Store: openai.Bool(false)}),
+		openaiprovider.ResponsesIncludeReasoningEncryptedContent(false),
+	).Collect()
+	if err != nil {
+		t.Fatalf("error = %v", err)
+	}
+}
+
 func TestResponsesConversationId_AsConversationId_NonStreaming(t *testing.T) {
 	const input = `
             {