diff --git a/cmd/verifyexamples/examples.go b/cmd/verifyexamples/examples.go
index fb3394cd..7ff7134c 100644
--- a/cmd/verifyexamples/examples.go
+++ b/cmd/verifyexamples/examples.go
@@ -812,6 +812,17 @@ var agentsExamples = []ExampleDefinition{
 			"The output should not contain error messages or stack traces.",
 		},
 	},
+	{
+		Name:                         "02_agents_harness_step03_agentmode",
+		ProjectPath:                  "examples/02-agents/harness/step03_agentmode",
+		RequiredEnvironmentVariables: []string{"FOUNDRY_PROJECT_ENDPOINT"},
+		OptionalEnvironmentVariables: []string{"FOUNDRY_MODEL"},
+		ExpectedOutputDescription: []string{
+			"The output should show a first run where the agent plans steps for a study group, then a second run where it executes them.",
+			"The two runs reflect the session switching from plan mode to execute mode between them.",
+			"The output should not contain error messages or stack traces.",
+		},
+	},
 	{
 		Name:                         "02_agents_harness_step01_loop",
 		ProjectPath:                  "examples/02-agents/harness/step01_loop",
diff --git a/examples/02-agents/harness/step03_agentmode/main.go b/examples/02-agents/harness/step03_agentmode/main.go
new file mode 100644
index 00000000..5995ae1b
--- /dev/null
+++ b/examples/02-agents/harness/step03_agentmode/main.go
@@ -0,0 +1,71 @@
+// Copyright (c) Microsoft. All rights reserved.
+
+package main
+
+import (
+	"context"
+
+	"github.com/microsoft/agent-framework-go/agent"
+	"github.com/microsoft/agent-framework-go/agent/harness/agentmode"
+	"github.com/microsoft/agent-framework-go/examples/internal/demo"
+	"github.com/microsoft/agent-framework-go/provider/foundryprovider"
+)
+
+var logger = demo.NewLogger(
+	"Agent Mode Harness",
+	"This sample shows how the agent-mode harness gives an agent a set of operating "+
+		"modes (plan / execute by default) plus mode_set/mode_get tools, and injects "+
+		"mode-specific instructions so the agent adjusts its behavior per mode.",
+	"Model", demo.FoundryModel,
+)
+
+func main() {
+	token := demo.FoundryTokenCredential()
+
+	// The agent-mode context provider injects the current mode's instructions at
+	// the start of each run. Because instructions are rebuilt per run, a single
+	// run only ever sees one mode; to demonstrate both, the example issues one
+	// run per mode and switches the session's mode in between. Keep a reference
+	// to the provider so we can drive that switch.
+	modeProvider := agentmode.New(agentmode.Config{})
+
+	a := foundryprovider.NewAgent(
+		demo.FoundryProjectEndpoint,
+		token,
+		foundryprovider.ModelDeployment(demo.FoundryModel),
+		foundryprovider.AgentConfig{
+			Instructions: "You are a capable assistant that follows the current operating mode's guidance.",
+			Config: agent.Config{
+				Name:             "ModeAssistant",
+				Middlewares:      []agent.Middleware{logger}, // for logging agent interactions
+				ContextProviders: []agent.ContextProvider{modeProvider},
+			},
+		},
+	)
+
+	ctx := context.Background()
+	session, err := a.CreateSession(ctx)
+	if err != nil {
+		demo.Panic(err)
+	}
+
+	// Run 1: the session starts in "plan" mode, so the provider injects the
+	// plan-mode instructions and the agent plans the work.
+	planResp, err := a.RunText(ctx,
+		"I want to organize a small study group. Plan the steps before doing anything.",
+		agent.WithSession(session)).Collect()
+	demo.Response(planResp, err)
+
+	// Switch the session to "execute" mode. The next run rebuilds the injected
+	// instructions from this mode, so the agent now carries out the plan.
+	if err := modeProvider.SetMode(session, "execute", false); err != nil {
+		demo.Panic(err)
+	}
+
+	// Run 2: now in "execute" mode, the provider injects the execute-mode
+	// instructions and the agent carries out the planned steps.
+	execResp, err := a.RunText(ctx,
+		"Now execute the plan you just made.",
+		agent.WithSession(session)).Collect()
+	demo.Response(execResp, err)
+}