diff --git a/cli/azd/.vscode/cspell.yaml b/cli/azd/.vscode/cspell.yaml
index f0a25445ce5..3a53d765055 100644
--- a/cli/azd/.vscode/cspell.yaml
+++ b/cli/azd/.vscode/cspell.yaml
@@ -177,6 +177,9 @@ dictionaryDefinitions:
 dictionaries:
   - azdProjectDictionary
 overrides:
+  - filename: "**/internal/agent/types.go"
+    words:
+      - billionized
   - filename: internal/appdetect/aspire_polyglot.go
     words:
       - pylock
diff --git a/cli/azd/cmd/telemetry_test.go b/cli/azd/cmd/telemetry_test.go
index cea6bf3b082..60c6cea35ff 100644
--- a/cli/azd/cmd/telemetry_test.go
+++ b/cli/azd/cmd/telemetry_test.go
@@ -132,6 +132,7 @@ func TestTelemetryFieldConstants(t *testing.T) {
 
 		measurementFields := []fields.AttributeKey{
 			fields.AgentFixAttempts,
+			fields.CopilotMessageAICredits,
 			fields.ExeGraphDeployConcurrencyKey,
 			fields.ExeGraphMaxConcurrencyKey,
 			fields.ExeGraphPackageConcurrencyKey,
diff --git a/cli/azd/docs/extensions/extension-framework.md b/cli/azd/docs/extensions/extension-framework.md
index c25b4736b78..d6102632b8e 100644
--- a/cli/azd/docs/extensions/extension-framework.md
+++ b/cli/azd/docs/extensions/extension-framework.md
@@ -3128,8 +3128,9 @@ Returns cumulative usage metrics cached for a session.
     - `input_tokens` (double): Total input tokens consumed
     - `output_tokens` (double): Total output tokens consumed
     - `total_tokens` (double): Sum of input + output tokens
-    - `billing_rate` (double): Per-request cost multiplier (e.g., 1.0x, 2.0x)
-    - `premium_requests` (double): Number of premium requests used
+    - `billing_rate` (double, deprecated): Legacy per-request cost multiplier; use `ai_credits` instead
+    - `premium_requests` (double, deprecated): Legacy premium request count; use `ai_credits` instead
+    - `ai_credits` (double): Total AI credits consumed
     - `duration_ms` (double): Total API duration in milliseconds
 
 #### GetFileChanges
@@ -3223,8 +3224,8 @@ metricsResp, err := copilot.GetUsageMetrics(ctx, &v1beta.GetCopilotUsageMetricsR
 if err != nil {
     return fmt.Errorf("failed to get metrics: %w", err)
 }
-fmt.Printf("Total tokens: %.0f, Premium requests: %.0f\n",
-    metricsResp.Usage.TotalTokens, metricsResp.Usage.PremiumRequests)
+fmt.Printf("Total tokens: %.0f, AI credits: %.2f AIC\n",
+  metricsResp.Usage.TotalTokens, metricsResp.Usage.AiCredits)
 
 // Retrieve file changes
 changesResp, err := copilot.GetFileChanges(ctx, &v1beta.GetCopilotFileChangesRequest{
diff --git a/cli/azd/extensions/microsoft.azd.demo/internal/cmd/copilot.go b/cli/azd/extensions/microsoft.azd.demo/internal/cmd/copilot.go
index 50de2c77c02..04cf4e89ef3 100644
--- a/cli/azd/extensions/microsoft.azd.demo/internal/cmd/copilot.go
+++ b/cli/azd/extensions/microsoft.azd.demo/internal/cmd/copilot.go
@@ -279,7 +279,7 @@ func showCumulativeMetrics(
 	}
 
 	usage := metricsResp.Usage
-	if usage == nil || (usage.InputTokens == 0 && usage.OutputTokens == 0) {
+	if !hasUsageMetrics(usage) {
 		return
 	}
 
@@ -294,18 +294,20 @@ func showCumulativeMetrics(
 		color.HiBlackString("•"), formatTokens(usage.OutputTokens))
 	fmt.Printf("  %s Total tokens:     %s\n",
 		color.HiBlackString("•"), formatTokens(usage.TotalTokens))
-	if usage.BillingRate > 0 {
-		fmt.Printf("  %s Billing rate:     %.0fx per request\n",
-			color.HiBlackString("•"), usage.BillingRate)
+	if usage.AiCredits > 0 {
+		fmt.Printf("  %s AI credits:       %s\n",
+			color.HiBlackString("•"), fmt.Sprintf("%.2f", usage.AiCredits))
 	}
-	fmt.Printf("  %s Premium requests: %.0f\n",
-		color.HiBlackString("•"), usage.PremiumRequests)
 	if usage.DurationMs > 0 {
 		fmt.Printf("  %s API duration:     %s\n",
 			color.HiBlackString("•"), formatDuration(usage.DurationMs))
 	}
 }
 
+func hasUsageMetrics(usage *v1beta.CopilotUsageMetrics) bool {
+	return usage != nil && (usage.InputTokens != 0 || usage.OutputTokens != 0 || usage.AiCredits != 0)
+}
+
 // showFileChanges displays accumulated file changes from the session.
 func showFileChanges(
 	ctx context.Context,
diff --git a/cli/azd/extensions/microsoft.azd.demo/internal/cmd/copilot_test.go b/cli/azd/extensions/microsoft.azd.demo/internal/cmd/copilot_test.go
new file mode 100644
index 00000000000..f5ed4d3ef82
--- /dev/null
+++ b/cli/azd/extensions/microsoft.azd.demo/internal/cmd/copilot_test.go
@@ -0,0 +1,35 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+package cmd
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/require"
+
+	v1beta "github.com/azure/azure-dev/cli/azd/pkg/azdext/contracts/v1beta"
+)
+
+func TestHasUsageMetrics(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name     string
+		usage    *v1beta.CopilotUsageMetrics
+		expected bool
+	}{
+		{name: "Nil", expected: false},
+		{name: "Empty", usage: &v1beta.CopilotUsageMetrics{}, expected: false},
+		{name: "InputTokens", usage: &v1beta.CopilotUsageMetrics{InputTokens: 1}, expected: true},
+		{name: "OutputTokens", usage: &v1beta.CopilotUsageMetrics{OutputTokens: 1}, expected: true},
+		{name: "AICredits", usage: &v1beta.CopilotUsageMetrics{AiCredits: 0.25}, expected: true},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			t.Parallel()
+			require.Equal(t, test.expected, hasUsageMetrics(test.usage))
+		})
+	}
+}
diff --git a/cli/azd/grpc/proto/azd/extensions/v1/ai_model.proto b/cli/azd/grpc/proto/azd/extensions/v1/ai_model.proto
index 185bb904e27..ee856756ac9 100644
--- a/cli/azd/grpc/proto/azd/extensions/v1/ai_model.proto
+++ b/cli/azd/grpc/proto/azd/extensions/v1/ai_model.proto
@@ -40,7 +40,8 @@ service AiModelService {
 message AiModel {
   string name = 1;                                // e.g. "gpt-4o"
   string format = 2;                              // e.g. "OpenAI"
-  string lifecycle_status = 3 [deprecated = true]; // deprecated; always empty; use AiModelVersion.lifecycle_status
+  // Always empty. Use AiModelVersion.lifecycle_status instead.
+  string lifecycle_status = 3 [deprecated = true];
   repeated string capabilities = 4;               // e.g. ["chat", "embeddings"]
   repeated AiModelVersion versions = 5;
   repeated string locations = 6;                  // canonical locations where available
diff --git a/cli/azd/grpc/proto/azd/extensions/v1beta/copilot.proto b/cli/azd/grpc/proto/azd/extensions/v1beta/copilot.proto
index d0104cc1b6d..c9a47403f73 100644
--- a/cli/azd/grpc/proto/azd/extensions/v1beta/copilot.proto
+++ b/cli/azd/grpc/proto/azd/extensions/v1beta/copilot.proto
@@ -118,9 +118,12 @@ message CopilotUsageMetrics {
   double input_tokens = 2;      // Total input tokens consumed.
   double output_tokens = 3;     // Total output tokens consumed.
   double total_tokens = 4;      // Sum of input + output tokens.
-  double billing_rate = 5;      // Per-request cost multiplier (e.g., 1.0x, 2.0x).
-  double premium_requests = 6;  // Number of premium requests used.
+  // Legacy per-request cost multiplier. Use ai_credits instead.
+  double billing_rate = 5 [deprecated = true];
+  // Legacy premium request count. Use ai_credits instead.
+  double premium_requests = 6 [deprecated = true];
   double duration_ms = 7;       // Total API duration in milliseconds.
+  double ai_credits = 8;        // Total AI credits consumed.
 }
 
 // --- File change messages ---
diff --git a/cli/azd/internal/agent/copilot/copilot_sdk_e2e_test.go b/cli/azd/internal/agent/copilot/copilot_sdk_e2e_test.go
index b540428903c..707f3fec5a5 100644
--- a/cli/azd/internal/agent/copilot/copilot_sdk_e2e_test.go
+++ b/cli/azd/internal/agent/copilot/copilot_sdk_e2e_test.go
@@ -1,19 +1,25 @@
 // Copyright (c) Microsoft Corporation. All rights reserved.
 // Licensed under the MIT License.
 
-package copilot
+package copilot_test
 
 import (
 	"context"
+	"encoding/json"
 	"fmt"
 	"net/http"
 	"os"
+	"slices"
+	"strings"
+	"sync"
 	"testing"
 	"time"
 
 	copilot "github.com/github/copilot-sdk/go"
 	"github.com/stretchr/testify/require"
 
+	"github.com/azure/azure-dev/cli/azd/internal/agent"
+	agentcopilot "github.com/azure/azure-dev/cli/azd/internal/agent/copilot"
 	"github.com/azure/azure-dev/cli/azd/test/mocks/mockinput"
 )
 
@@ -39,8 +45,8 @@ func TestCopilotSDK_E2E(t *testing.T) {
 	t.Setenv("AZD_COPILOT_CLI_PATH", "")
 
 	// 1. Download the pinned CLI and start it through azd's client manager.
-	cli := NewCopilotCLI(mockinput.NewMockConsole(), nil, http.DefaultClient)
-	clientManager := NewCopilotClientManager(&CopilotClientOptions{
+	cli := agentcopilot.NewCopilotCLI(mockinput.NewMockConsole(), nil, http.DefaultClient)
+	clientManager := agentcopilot.NewCopilotClientManager(&agentcopilot.CopilotClientOptions{
 		LogLevel: "error",
 	}, cli)
 
@@ -58,7 +64,7 @@ func TestCopilotSDK_E2E(t *testing.T) {
 	// 2. Check auth
 	auth, err := clientManager.GetAuthStatus(ctx)
 	require.NoError(t, err)
-	t.Logf("Auth: authenticated=%v, login=%v", auth.IsAuthenticated, auth.Login)
+	t.Logf("Auth: authenticated=%v", auth.IsAuthenticated)
 	require.True(t, auth.IsAuthenticated, "not authenticated with GitHub Copilot")
 
 	// 3. List models
@@ -81,7 +87,7 @@ func TestCopilotSDK_E2E(t *testing.T) {
 		OnPermissionRequest: copilot.PermissionHandler.ApproveAll,
 	})
 	require.NoError(t, err, "CreateSession failed")
-	t.Logf("Session created: %s", session.WorkspacePath())
+	t.Log("Session created")
 	defer func() {
 		if disconnectErr := session.Disconnect(); disconnectErr != nil {
 			t.Logf("session.Destroy error: %v", disconnectErr)
@@ -89,50 +95,174 @@ func TestCopilotSDK_E2E(t *testing.T) {
 	}()
 
 	// 5. Collect events
-	var events []copilot.SessionEvent
+	collector := agent.NewHeadlessCollector()
+	captured := &capturedEvents{}
 	unsubscribe := session.On(func(event copilot.SessionEvent) {
-		events = append(events, event)
 		t.Logf("Event: type=%s", event.Type())
+		aiuFields, err := formatAIUFields(event)
+		captured.Add(event, err)
+		if aiuFields != "" {
+			t.Logf("AIU event: type=%s %s", event.Type(), aiuFields)
+		}
+		collector.HandleEvent(event)
 	})
 	defer unsubscribe()
 
-	// 6. Send message and wait for response
-	t.Log("Sending prompt...")
-	response, err := session.SendAndWait(ctx, copilot.MessageOptions{
+	// 6. Send two messages in the same session and wait for each turn to become idle.
+	t.Log("Sending first prompt...")
+	firstEvent := captured.Len()
+	firstResponse, err := session.SendAndWait(ctx, copilot.MessageOptions{
 		Prompt: "What is 2+2? Reply with just the number.",
 	})
-	require.NoError(t, err, "SendAndWait failed")
+	require.NoError(t, err, "first SendAndWait failed")
+	require.NoError(t, collector.WaitForIdle(ctx), "collector did not observe first session idle")
+	require.NoError(t, captured.Err())
+	requireResponseContains(t, firstResponse, captured.Since(firstEvent), "4")
+
+	firstUsage := collector.GetUsageMetrics()
+	require.Positive(t, firstUsage.AICredits, "expected positive AI credit usage after first turn")
+
+	t.Log("Sending follow-up prompt...")
+	secondEvent := captured.Len()
+	secondResponse, err := session.SendAndWait(ctx, copilot.MessageOptions{
+		Prompt: "Add 3 to the number you answered in the previous turn. Reply with just the result.",
+	})
+	require.NoError(t, err, "second SendAndWait failed")
+	require.NoError(t, collector.WaitForIdle(ctx), "collector did not observe second session idle")
+	require.NoError(t, captured.Err())
+	requireResponseContains(t, secondResponse, captured.Since(secondEvent), "7")
+
+	usage := collector.GetUsageMetrics()
+	require.Greater(t, usage.InputTokens, firstUsage.InputTokens)
+	require.Greater(t, usage.OutputTokens, firstUsage.OutputTokens)
+	require.Greater(t, usage.AICredits, firstUsage.AICredits)
+	t.Logf("Usage: input=%v output=%v AI credits=%v", usage.InputTokens, usage.OutputTokens, usage.AICredits)
 
 	// 7. Validate response
-	t.Logf("Received %d events total", len(events))
+	t.Logf("Received %d events total", captured.Len())
+}
+
+type capturedEvents struct {
+	mu     sync.Mutex
+	events []copilot.SessionEvent
+	err    error
+}
+
+func (c *capturedEvents) Add(event copilot.SessionEvent, err error) {
+	c.mu.Lock()
+	defer c.mu.Unlock()
+
+	c.events = append(c.events, event)
+	if c.err == nil {
+		c.err = err
+	}
+}
+
+func (c *capturedEvents) Len() int {
+	c.mu.Lock()
+	defer c.mu.Unlock()
+
+	return len(c.events)
+}
+
+func (c *capturedEvents) Since(index int) []copilot.SessionEvent {
+	c.mu.Lock()
+	defer c.mu.Unlock()
+
+	return slices.Clone(c.events[index:])
+}
+
+func (c *capturedEvents) Err() error {
+	c.mu.Lock()
+	defer c.mu.Unlock()
+
+	return c.err
+}
+
+func requireResponseContains(
+	t *testing.T,
+	response *copilot.SessionEvent,
+	events []copilot.SessionEvent,
+	expected string,
+) {
+	t.Helper()
+
 	if response != nil {
 		data, ok := response.Data.(*copilot.AssistantMessageData)
 		require.True(t, ok, "expected response.Data to be *copilot.AssistantMessageData, got %T", response.Data)
 		t.Logf("Response content: %s", data.Content)
-		require.Contains(t, data.Content, "4",
-			"expected response to contain '4'")
+		require.Contains(t, data.Content, expected)
 	} else {
-		// If SendAndWait returned nil, check events for assistant message
 		var found bool
-		for _, e := range events {
-			if e.Type() == copilot.SessionEventTypeAssistantMessage {
-				if data, ok := e.Data.(*copilot.AssistantMessageData); ok {
+		for _, event := range events {
+			if event.Type() == copilot.SessionEventTypeAssistantMessage {
+				if data, ok := event.Data.(*copilot.AssistantMessageData); ok {
 					t.Logf("Found assistant message in events: %s", data.Content)
-					found = true
-					break
+					if strings.Contains(data.Content, expected) {
+						found = true
+						break
+					}
 				}
 			}
 		}
 		if !found {
-			// Log all event types for debugging
-			for _, e := range events {
+			for _, event := range events {
 				detail := ""
-				if data, ok := e.Data.(*copilot.AssistantMessageData); ok {
+				if data, ok := event.Data.(*copilot.AssistantMessageData); ok {
 					detail = fmt.Sprintf(" content=%s", truncateForLog(data.Content, 100))
 				}
-				t.Logf("  event: type=%s%s", e.Type(), detail)
+				t.Logf("  event: type=%s%s", event.Type(), detail)
+			}
+			t.Fatalf("no assistant message containing %q received", expected)
+		}
+	}
+}
+
+func formatAIUFields(event copilot.SessionEvent) (string, error) {
+	encoded, err := json.Marshal(event.Data)
+	if err != nil {
+		return "", fmt.Errorf("marshaling %s event data: %w", event.Type(), err)
+	}
+
+	var data any
+	if err := json.Unmarshal(encoded, &data); err != nil {
+		return "", fmt.Errorf("unmarshaling %s event data: %w", event.Type(), err)
+	}
+
+	var fields []string
+	collectAIUFields(data, "data", &fields)
+	if len(fields) == 0 {
+		return "", nil
+	}
+	if usage, ok := event.Data.(*copilot.AssistantUsageData); ok {
+		if usage.InputTokens != nil {
+			fields = append(fields, fmt.Sprintf("data.inputTokens=%d", *usage.InputTokens))
+		}
+		if usage.OutputTokens != nil {
+			fields = append(fields, fmt.Sprintf("data.outputTokens=%d", *usage.OutputTokens))
+		}
+	}
+
+	slices.Sort(fields)
+	return strings.Join(fields, ", "), nil
+}
+
+func collectAIUFields(value any, path string, fields *[]string) {
+	switch value := value.(type) {
+	case map[string]any:
+		for name, child := range value {
+			childPath := path + "." + name
+			if name == "totalNanoAiu" {
+				if number, ok := child.(float64); ok {
+					*fields = append(*fields, fmt.Sprintf("%s=%g", childPath, number))
+				}
+				continue
 			}
-			t.Fatal("no assistant message received")
+			collectAIUFields(child, childPath, fields)
+		}
+	case []any:
+		for _, child := range value {
+			collectAIUFields(child, path+"[]", fields)
 		}
 	}
 }
diff --git a/cli/azd/internal/agent/copilot_agent.go b/cli/azd/internal/agent/copilot_agent.go
index 6d0b9cdd518..a252bdc1ced 100644
--- a/cli/azd/internal/agent/copilot_agent.go
+++ b/cli/azd/internal/agent/copilot_agent.go
@@ -56,7 +56,7 @@ type CopilotAgent struct {
 
 	// Runtime state
 	clientStarted          bool
-	session                *copilot.Session
+	session                copilotSession
 	sessionID              string
 	activeCtx              atomic.Pointer[context.Context] // current SendMessage context for SDK callbacks
 	display                *AgentDisplay                   // last display for usage metrics (interactive mode)
@@ -76,6 +76,12 @@ type cleanupTask struct {
 	fn   func() error
 }
 
+type copilotSession interface {
+	On(copilot.SessionEventHandler) func()
+	Send(context.Context, copilot.MessageOptions) (string, error)
+	GetEvents(context.Context) ([]copilot.SessionEvent, error)
+}
+
 const (
 	selectAIModelMessage              = "Select AI model"
 	selectReasoningEffortLevelMessage = "Select reasoning effort level"
@@ -271,16 +277,21 @@ func (a *CopilotAgent) sendMessageInteractive(
 		return nil, fmt.Errorf("copilot agent error: %w", err)
 	}
 
-	if err := display.WaitForIdle(ctx); err != nil {
-		return nil, err
-	}
-
+	err = display.WaitForIdle(ctx)
 	turnUsage := display.GetUsageMetrics()
+
 	a.mu.Lock()
 	a.accumulateUsage(turnUsage)
-	turnFileChanges := a.collectFileChanges(watcher)
+	var turnFileChanges watch.FileChanges
+	if err == nil {
+		turnFileChanges = a.collectFileChanges(watcher)
+	}
 	a.mu.Unlock()
 
+	if err != nil {
+		return nil, err
+	}
+
 	return &AgentResult{
 		SessionID:   a.sessionID,
 		Usage:       turnUsage,
@@ -312,16 +323,21 @@ func (a *CopilotAgent) sendMessageHeadless(
 		return nil, fmt.Errorf("copilot agent error: %w", err)
 	}
 
-	if err := collector.WaitForIdle(ctx); err != nil {
-		return nil, err
-	}
-
+	err = collector.WaitForIdle(ctx)
 	turnUsage := collector.GetUsageMetrics()
+
 	a.mu.Lock()
 	a.accumulateUsage(turnUsage)
-	turnFileChanges := a.collectFileChanges(watcher)
+	var turnFileChanges watch.FileChanges
+	if err == nil {
+		turnFileChanges = a.collectFileChanges(watcher)
+	}
 	a.mu.Unlock()
 
+	if err != nil {
+		return nil, err
+	}
+
 	return &AgentResult{
 		SessionID:   a.sessionID,
 		Usage:       turnUsage,
@@ -333,6 +349,7 @@ func (a *CopilotAgent) sendMessageHeadless(
 func (a *CopilotAgent) accumulateUsage(turn UsageMetrics) {
 	a.cumulativeUsage.InputTokens += turn.InputTokens
 	a.cumulativeUsage.OutputTokens += turn.OutputTokens
+	a.cumulativeUsage.AICredits += turn.AICredits
 	a.cumulativeUsage.DurationMS += turn.DurationMS
 	a.cumulativeUsage.PremiumRequests += turn.PremiumRequests
 	// These are per-request values, not cumulative — use latest
@@ -405,8 +422,7 @@ func (a *CopilotAgent) Stop() error {
 		fields.CopilotMessageModel.String(a.cumulativeUsage.Model),
 		fields.CopilotMessageInputTokens.Float64(a.cumulativeUsage.InputTokens),
 		fields.CopilotMessageOutputTokens.Float64(a.cumulativeUsage.OutputTokens),
-		fields.CopilotMessageBillingRate.Float64(a.cumulativeUsage.BillingRate),
-		fields.CopilotMessagePremiumRequests.Float64(a.cumulativeUsage.PremiumRequests),
+		fields.CopilotMessageAICredits.Float64(a.cumulativeUsage.AICredits),
 		fields.CopilotMessageDurationMs.Float64(a.cumulativeUsage.DurationMS),
 		fields.CopilotConsentApprovedCount.Int(a.consentApprovedCount),
 		fields.CopilotConsentDeniedCount.Int(a.consentDeniedCount),
diff --git a/cli/azd/internal/agent/copilot_agent_test.go b/cli/azd/internal/agent/copilot_agent_test.go
index b9ce2aa198e..57ad5ab16bf 100644
--- a/cli/azd/internal/agent/copilot_agent_test.go
+++ b/cli/azd/internal/agent/copilot_agent_test.go
@@ -14,11 +14,112 @@ import (
 	"github.com/stretchr/testify/require"
 
 	agentcopilot "github.com/azure/azure-dev/cli/azd/internal/agent/copilot"
+	"github.com/azure/azure-dev/cli/azd/internal/tracing"
+	"github.com/azure/azure-dev/cli/azd/internal/tracing/fields"
 	"github.com/azure/azure-dev/cli/azd/pkg/config"
 	"github.com/azure/azure-dev/cli/azd/pkg/input"
 	"github.com/azure/azure-dev/cli/azd/test/mocks"
 )
 
+type fakeCopilotSession struct {
+	handler copilot.SessionEventHandler
+	send    func(context.Context, copilot.MessageOptions) (string, error)
+}
+
+func (s *fakeCopilotSession) On(handler copilot.SessionEventHandler) func() {
+	s.handler = handler
+	return func() { s.handler = nil }
+}
+
+func (s *fakeCopilotSession) Send(ctx context.Context, options copilot.MessageOptions) (string, error) {
+	return s.send(ctx, options)
+}
+
+func (s *fakeCopilotSession) GetEvents(context.Context) ([]copilot.SessionEvent, error) {
+	return nil, nil
+}
+
+func TestCopilotAgentStopRecordsAICredits(t *testing.T) {
+	tracing.ResetUsageAttributesForTest()
+	t.Cleanup(tracing.ResetUsageAttributesForTest)
+
+	agent := &CopilotAgent{
+		cumulativeUsage: UsageMetrics{AICredits: 1.25},
+	}
+	require.NoError(t, agent.Stop())
+
+	for _, attr := range tracing.GetUsageAttributes() {
+		if attr.Key == fields.CopilotMessageAICredits.Key {
+			require.Equal(t, 1.25, attr.Value.AsFloat64())
+			return
+		}
+	}
+	require.Fail(t, "copilot AI-credit usage attribute was not recorded")
+}
+
+func TestCopilotAgentAccumulateUsage(t *testing.T) {
+	agent := &CopilotAgent{}
+
+	agent.accumulateUsage(UsageMetrics{
+		Model:           "gpt-4o",
+		InputTokens:     100,
+		OutputTokens:    50,
+		AICredits:       0.25,
+		BillingRate:     1.5,
+		PremiumRequests: 2,
+		DurationMS:      500,
+	})
+	agent.accumulateUsage(UsageMetrics{
+		Model:           "gpt-4.1",
+		InputTokens:     200,
+		OutputTokens:    75,
+		AICredits:       0.5,
+		BillingRate:     2,
+		PremiumRequests: 3,
+		DurationMS:      1000,
+	})
+
+	usage := agent.GetMetrics().Usage
+	require.Equal(t, UsageMetrics{
+		Model:           "gpt-4.1",
+		InputTokens:     300,
+		OutputTokens:    125,
+		AICredits:       0.75,
+		BillingRate:     2,
+		PremiumRequests: 5,
+		DurationMS:      1500,
+	}, usage)
+}
+
+func TestCopilotAgentSendMessageHeadlessRecordsUsageOnCancellation(t *testing.T) {
+	ctx, cancel := context.WithCancel(t.Context())
+	session := &fakeCopilotSession{}
+	session.send = func(context.Context, copilot.MessageOptions) (string, error) {
+		session.handler(copilot.SessionEvent{
+			Data: &copilot.AssistantUsageData{
+				InputTokens:  new(int64(100)),
+				OutputTokens: new(int64(50)),
+				CopilotUsage: &copilot.AssistantUsageCopilotUsage{TotalNanoAiu: 250_000_000},
+				Duration:     new(int64(1000)),
+			},
+		})
+		cancel()
+		return "", nil
+	}
+
+	agent := &CopilotAgent{session: session, sessionID: "test-session"}
+	result, err := agent.sendMessageHeadless(ctx, "test", AgentModeAutopilot)
+
+	require.ErrorIs(t, err, context.Canceled)
+	require.Nil(t, result)
+	require.Equal(t, UsageMetrics{
+		InputTokens:  100,
+		OutputTokens: 50,
+		AICredits:    0.25,
+		DurationMS:   1000,
+	}, agent.GetMetrics().Usage)
+}
+
 func TestCopilotAgentPromptModelAndReasoning(t *testing.T) {
 	t.Run("ModelConfigurationAlreadyInConfig", func(t *testing.T) {
 		runCopilotAgentTest(t, copilotAgentTestArgs{
diff --git a/cli/azd/internal/agent/display.go b/cli/azd/internal/agent/display.go
index 6131a62fc2c..bf8689a5601 100644
--- a/cli/azd/internal/agent/display.go
+++ b/cli/azd/internal/agent/display.go
@@ -53,6 +53,7 @@ type AgentDisplay struct {
 	// Usage metrics — accumulated from assistant.usage events
 	totalInputTokens  float64
 	totalOutputTokens float64
+	totalNanoAiu      float64
 	billingRate       float64
 	totalDurationMS   float64
 	premiumRequests   float64
@@ -332,6 +333,9 @@ func (d *AgentDisplay) HandleEvent(event copilot.SessionEvent) {
 			if data.OutputTokens != nil {
 				d.totalOutputTokens += float64(*data.OutputTokens)
 			}
+			if data.CopilotUsage != nil {
+				d.totalNanoAiu += data.CopilotUsage.TotalNanoAiu
+			}
 			if data.Cost != nil {
 				d.billingRate = *data.Cost // per-request multiplier, not cumulative
 			}
@@ -643,6 +647,7 @@ func (d *AgentDisplay) GetUsageMetrics() UsageMetrics {
 		Model:           d.lastModel,
 		InputTokens:     d.totalInputTokens,
 		OutputTokens:    d.totalOutputTokens,
+		AICredits:       nanoAiuToCredits(d.totalNanoAiu),
 		BillingRate:     d.billingRate,
 		PremiumRequests: d.premiumRequests,
 		DurationMS:      d.totalDurationMS,
diff --git a/cli/azd/internal/agent/display_test.go b/cli/azd/internal/agent/display_test.go
index 241e8e53ea6..9044393898a 100644
--- a/cli/azd/internal/agent/display_test.go
+++ b/cli/azd/internal/agent/display_test.go
@@ -111,15 +111,19 @@ func TestGetUsageMetrics(t *testing.T) {
 	outputTokens1 := int64(500)
 	cost1 := float64(1.0)
 	duration1 := int64(5000)
+	nanoAiu1 := 1_500_000_000.0
 
 	// Simulate usage events
 	d.HandleEvent(copilot.SessionEvent{
 		Data: &copilot.AssistantUsageData{
 			InputTokens:  &inputTokens1,
 			OutputTokens: &outputTokens1,
-			Cost:         &cost1,
-			Duration:     &duration1,
-			Model:        "gpt-4.1",
+			CopilotUsage: &copilot.AssistantUsageCopilotUsage{
+				TotalNanoAiu: nanoAiu1,
+			},
+			Cost:     &cost1,
+			Duration: &duration1,
+			Model:    "gpt-4.1",
 		},
 	})
 
@@ -127,19 +131,27 @@ func TestGetUsageMetrics(t *testing.T) {
 	outputTokens2 := int64(800)
 	cost2 := float64(1.0)
 	duration2 := int64(3000)
+	nanoAiu2 := 1_000_000_000.0
 
 	d.HandleEvent(copilot.SessionEvent{
 		Data: &copilot.AssistantUsageData{
 			InputTokens:  &inputTokens2,
 			OutputTokens: &outputTokens2,
-			Cost:         &cost2,
-			Duration:     &duration2,
+			CopilotUsage: &copilot.AssistantUsageCopilotUsage{
+				TotalNanoAiu: nanoAiu2,
+			},
+			Cost:     &cost2,
+			Duration: &duration2,
 		},
 	})
+	d.HandleEvent(copilot.SessionEvent{Data: &copilot.AssistantMessageData{}})
+	d.HandleEvent(copilot.SessionEvent{Data: &copilot.SessionIdleData{}})
+	require.NoError(t, d.WaitForIdle(t.Context()))
 
 	metrics := d.GetUsageMetrics()
 	require.Equal(t, float64(3000), metrics.InputTokens)
 	require.Equal(t, float64(1300), metrics.OutputTokens)
+	require.Equal(t, 2.5, metrics.AICredits)
 	require.Equal(t, float64(1.0), metrics.BillingRate) // last value, not sum
 	require.Equal(t, float64(8000), metrics.DurationMS)
 	require.Equal(t, "gpt-4.1", metrics.Model)
diff --git a/cli/azd/internal/agent/headless_collector.go b/cli/azd/internal/agent/headless_collector.go
index 5a0263563fa..d58d84b4dda 100644
--- a/cli/azd/internal/agent/headless_collector.go
+++ b/cli/azd/internal/agent/headless_collector.go
@@ -20,6 +20,7 @@ type HeadlessCollector struct {
 	// Usage metrics — accumulated from assistant.usage events
 	totalInputTokens  float64
 	totalOutputTokens float64
+	totalNanoAiu      float64
 	billingRate       float64
 	totalDurationMS   float64
 	premiumRequests   float64
@@ -68,6 +69,9 @@ func (h *HeadlessCollector) HandleEvent(event copilot.SessionEvent) {
 			if data.OutputTokens != nil {
 				h.totalOutputTokens += float64(*data.OutputTokens)
 			}
+			if data.CopilotUsage != nil {
+				h.totalNanoAiu += data.CopilotUsage.TotalNanoAiu
+			}
 			if data.Cost != nil {
 				h.billingRate = *data.Cost
 			}
@@ -128,6 +132,7 @@ func (h *HeadlessCollector) GetUsageMetrics() UsageMetrics {
 		Model:           h.lastModel,
 		InputTokens:     h.totalInputTokens,
 		OutputTokens:    h.totalOutputTokens,
+		AICredits:       nanoAiuToCredits(h.totalNanoAiu),
 		BillingRate:     h.billingRate,
 		PremiumRequests: h.premiumRequests,
 		DurationMS:      h.totalDurationMS,
diff --git a/cli/azd/internal/agent/headless_collector_test.go b/cli/azd/internal/agent/headless_collector_test.go
index 4c3c4cb925a..feb21750302 100644
--- a/cli/azd/internal/agent/headless_collector_test.go
+++ b/cli/azd/internal/agent/headless_collector_test.go
@@ -4,6 +4,8 @@
 package agent
 
 import (
+	"encoding/json"
+	"os"
 	"testing"
 
 	copilot "github.com/github/copilot-sdk/go"
@@ -100,14 +102,39 @@ func TestHeadlessCollector_WaitForIdle_DeferredIdle(t *testing.T) {
 	require.NoError(t, err)
 }
 
-func TestHeadlessCollector_PremiumRequests(t *testing.T) {
+func TestHeadlessCollector_ReplaysCapturedAIUEvents(t *testing.T) {
 	t.Parallel()
-	collector := NewHeadlessCollector()
 
-	collector.HandleEvent(copilot.SessionEvent{
-		Data: &copilot.SessionShutdownData{TotalPremiumRequests: new(5.0)},
-	})
+	contents, err := os.ReadFile("testdata/copilot_usage_events.json")
+	require.NoError(t, err)
 
-	usage := collector.GetUsageMetrics()
-	require.Equal(t, float64(5), usage.PremiumRequests)
+	var events []copilot.SessionEvent
+	require.NoError(t, json.Unmarshal(contents, &events))
+
+	collector := NewHeadlessCollector()
+	var usageByTurn []UsageMetrics
+	var checkpointNanoAiu []float64
+	for _, event := range events {
+		collector.HandleEvent(event)
+		if checkpoint, ok := event.Data.(*copilot.SessionUsageCheckpointData); ok {
+			checkpointNanoAiu = append(checkpointNanoAiu, checkpoint.TotalNanoAiu)
+		}
+		if event.Type() == copilot.SessionEventTypeSessionIdle {
+			require.NoError(t, collector.WaitForIdle(t.Context()))
+			usageByTurn = append(usageByTurn, collector.GetUsageMetrics())
+		}
+	}
+
+	require.Len(t, usageByTurn, 2)
+	require.Equal(t, 9969.0, usageByTurn[0].InputTokens)
+	require.Equal(t, 3.0, usageByTurn[0].OutputTokens)
+	require.Equal(t, 2.49515, usageByTurn[0].AICredits)
+	require.Equal(t, 20053.0, usageByTurn[1].InputTokens)
+	require.Equal(t, 6.0, usageByTurn[1].OutputTokens)
+	require.Equal(t, 2.72664, usageByTurn[1].AICredits)
+	require.NotContains(t, usageByTurn[1].String(), "Premium requests")
+
+	require.Len(t, checkpointNanoAiu, 2)
+	require.Equal(t, usageByTurn[0].AICredits, nanoAiuToCredits(checkpointNanoAiu[0]))
+	require.Equal(t, usageByTurn[1].AICredits, nanoAiuToCredits(checkpointNanoAiu[1]))
 }
diff --git a/cli/azd/internal/agent/testdata/copilot_usage_events.json b/cli/azd/internal/agent/testdata/copilot_usage_events.json
new file mode 100644
index 00000000000..b55bd7a9b3d
--- /dev/null
+++ b/cli/azd/internal/agent/testdata/copilot_usage_events.json
@@ -0,0 +1,74 @@
+[
+    {
+        "type": "assistant.turn_start",
+        "data": {}
+    },
+    {
+        "type": "assistant.usage",
+        "data": {
+            "inputTokens": 9969,
+            "outputTokens": 3,
+            "copilotUsage": {
+                "totalNanoAiu": 2495150000
+            }
+        }
+    },
+    {
+        "type": "assistant.message",
+        "data": {}
+    },
+    {
+        "type": "assistant.turn_end",
+        "data": {}
+    },
+    {
+        "type": "session.usage_checkpoint",
+        "data": {
+            "totalNanoAiu": 2495150000
+        }
+    },
+    {
+        "type": "assistant.idle",
+        "data": {}
+    },
+    {
+        "type": "session.idle",
+        "data": {}
+    },
+    {
+        "type": "assistant.turn_start",
+        "data": {}
+    },
+    {
+        "type": "assistant.usage",
+        "data": {
+            "inputTokens": 10084,
+            "outputTokens": 3,
+            "copilotUsage": {
+                "totalNanoAiu": 231490000
+            }
+        }
+    },
+    {
+        "type": "assistant.message",
+        "data": {}
+    },
+    {
+        "type": "assistant.turn_end",
+        "data": {}
+    },
+    {
+        "type": "session.usage_checkpoint",
+        "data": {
+            "totalNanoAiu": 2726640000
+        }
+    },
+    {
+        "type": "assistant.idle",
+        "data": {}
+    },
+    {
+        "type": "session.idle",
+        "data": {}
+    }
+]
diff --git a/cli/azd/internal/agent/types.go b/cli/azd/internal/agent/types.go
index 18a2c9544b7..8f974779dff 100644
--- a/cli/azd/internal/agent/types.go
+++ b/cli/azd/internal/agent/types.go
@@ -26,10 +26,17 @@ type AgentResult struct {
 
 // UsageMetrics tracks resource consumption for an agent session.
 type UsageMetrics struct {
-	Model           string
-	InputTokens     float64
-	OutputTokens    float64
-	BillingRate     float64 // per-request cost multiplier (e.g., 1.0x, 2.0x)
+	Model        string
+	InputTokens  float64
+	OutputTokens float64
+
+	// AICredits is the de-billionized AI-credit value derived from the SDK's raw
+	// nano-AIU totals. It should match what you see as "AIC" in other Copilots.
+	AICredits float64
+
+	// BillingRate is deprecated legacy metadata. Prefer AICredits for billing.
+	BillingRate float64
+	// PremiumRequests is deprecated legacy metadata. Prefer AICredits for billing.
 	PremiumRequests float64
 	DurationMS      float64
 }
@@ -39,9 +46,13 @@ func (u UsageMetrics) TotalTokens() float64 {
 	return u.InputTokens + u.OutputTokens
 }
 
+func nanoAiuToCredits(nanoAiu float64) float64 {
+	return nanoAiu / 1_000_000_000
+}
+
 // String returns a multi-line formatted string for display.
 func (u UsageMetrics) String() string {
-	if u.InputTokens == 0 && u.OutputTokens == 0 {
+	if u.InputTokens == 0 && u.OutputTokens == 0 && u.AICredits == 0 {
 		return ""
 	}
 
@@ -55,11 +66,10 @@ func (u UsageMetrics) String() string {
 	lines = append(lines, output.WithGrayFormat("  • Input tokens:     %s", formatTokenCount(u.InputTokens)))
 	lines = append(lines, output.WithGrayFormat("  • Output tokens:    %s", formatTokenCount(u.OutputTokens)))
 	lines = append(lines, output.WithGrayFormat("  • Total tokens:     %s", formatTokenCount(u.TotalTokens())))
-
-	if u.BillingRate > 0 {
-		lines = append(lines, output.WithGrayFormat("  • Billing rate:     %.0fx per request", u.BillingRate))
+	if u.AICredits > 0 {
+		lines = append(lines, output.WithGrayFormat("  • AI credits:       %.2f AIC", u.AICredits))
 	}
-	lines = append(lines, output.WithGrayFormat("  • Premium requests: %.0f", u.PremiumRequests))
+
 	if u.DurationMS > 0 {
 		seconds := u.DurationMS / 1000
 		if seconds >= 60 {
diff --git a/cli/azd/internal/agent/types_test.go b/cli/azd/internal/agent/types_test.go
index 67e0b4fb29c..e7962d39a35 100644
--- a/cli/azd/internal/agent/types_test.go
+++ b/cli/azd/internal/agent/types_test.go
@@ -37,16 +37,18 @@ func TestUsageMetrics_Format(t *testing.T) {
 		require.Contains(t, result, "claude-sonnet-4.5")
 	})
 
-	t.Run("WithCostAndPremium", func(t *testing.T) {
+	t.Run("LegacyBillingFieldsAreHidden", func(t *testing.T) {
 		u := UsageMetrics{
 			InputTokens:     50000,
 			OutputTokens:    20000,
+			AICredits:       0.25,
 			BillingRate:     2.0,
 			PremiumRequests: 15,
 		}
 		result := u.String()
-		require.Contains(t, result, "2x per request")
-		require.Contains(t, result, "15")
+		require.Contains(t, result, "AI credits:       0.25 AIC")
+		require.NotContains(t, result, "Premium requests")
+		require.NotContains(t, result, "2x per request")
 	})
 
 	t.Run("DurationSeconds", func(t *testing.T) {
diff --git a/cli/azd/internal/grpcserver/copilot_service.go b/cli/azd/internal/grpcserver/copilot_service.go
index d8cfe537a48..b477c32dae5 100644
--- a/cli/azd/internal/grpcserver/copilot_service.go
+++ b/cli/azd/internal/grpcserver/copilot_service.go
@@ -315,6 +315,7 @@ func convertUsageMetrics(usage agent.UsageMetrics) *azdext.CopilotUsageMetrics {
 		BillingRate:     usage.BillingRate,
 		PremiumRequests: usage.PremiumRequests,
 		DurationMs:      usage.DurationMS,
+		AiCredits:       usage.AICredits,
 	}
 }
 
diff --git a/cli/azd/internal/grpcserver/copilot_service_test.go b/cli/azd/internal/grpcserver/copilot_service_test.go
index 850f95da14d..70bd8ac9698 100644
--- a/cli/azd/internal/grpcserver/copilot_service_test.go
+++ b/cli/azd/internal/grpcserver/copilot_service_test.go
@@ -218,7 +218,7 @@ func TestCopilotService_GetUsageMetrics_ValidSession(t *testing.T) {
 	}, nil)
 	mockAgent.On("GetMetrics").Return(agent.AgentMetrics{
 		Usage: agent.UsageMetrics{
-			Model: "gpt-4o", InputTokens: 500, OutputTokens: 250, DurationMS: 3000,
+			Model: "gpt-4o", InputTokens: 500, OutputTokens: 250, AICredits: 1.25, DurationMS: 3000,
 		},
 	})
 
@@ -238,6 +238,7 @@ func TestCopilotService_GetUsageMetrics_ValidSession(t *testing.T) {
 	require.Equal(t, "gpt-4o", resp.Usage.Model)
 	require.Equal(t, float64(500), resp.Usage.InputTokens)
 	require.Equal(t, float64(250), resp.Usage.OutputTokens)
+	require.Equal(t, 1.25, resp.Usage.AiCredits)
 	require.Equal(t, float64(3000), resp.Usage.DurationMs)
 }
 
diff --git a/cli/azd/internal/grpcserver/prompt_service_test.go b/cli/azd/internal/grpcserver/prompt_service_test.go
index 52eca5df2d1..a0ae846e14e 100644
--- a/cli/azd/internal/grpcserver/prompt_service_test.go
+++ b/cli/azd/internal/grpcserver/prompt_service_test.go
@@ -1471,20 +1471,16 @@ func TestConvertFileChanges_WithChanges(t *testing.T) {
 func TestConvertUsageMetrics(t *testing.T) {
 	t.Parallel()
 	usage := agent.UsageMetrics{
-		Model:           "gpt-4o",
-		InputTokens:     100,
-		OutputTokens:    50,
-		BillingRate:     0.5,
-		PremiumRequests: 2,
-		DurationMS:      1500,
+		Model:        "gpt-4o",
+		InputTokens:  100,
+		OutputTokens: 50,
+		DurationMS:   1500,
 	}
 	result := convertUsageMetrics(usage)
 	require.Equal(t, "gpt-4o", result.Model)
 	require.Equal(t, float64(100), result.InputTokens)
 	require.Equal(t, float64(50), result.OutputTokens)
 	require.Equal(t, float64(150), result.TotalTokens) // 100 + 50
-	require.Equal(t, 0.5, result.BillingRate)
-	require.Equal(t, float64(2), result.PremiumRequests)
 	require.Equal(t, float64(1500), result.DurationMs)
 }
 
diff --git a/cli/azd/internal/telemetry/telemetry_test.go b/cli/azd/internal/telemetry/telemetry_test.go
index 48799d49cb9..687da333b00 100644
--- a/cli/azd/internal/telemetry/telemetry_test.go
+++ b/cli/azd/internal/telemetry/telemetry_test.go
@@ -15,6 +15,8 @@ import (
 )
 
 func TestGetTelemetrySystem(t *testing.T) {
+	t.Setenv("AZD_CONFIG_DIR", t.TempDir())
+
 	devEndpointConfig, err := appinsightsexporter.NewEndpointConfig(devConnectionString)
 	require.NoError(t, err)
 	prodEndpointConfig, err := appinsightsexporter.NewEndpointConfig(prodConnectionString)
@@ -100,7 +102,11 @@ func TestGetTelemetrySystem(t *testing.T) {
 }
 
 func TestTelemetrySystem_RunBackgroundUpload(t *testing.T) {
-	t.Parallel()
+	resetTelemetryForTest()
+	t.Cleanup(resetTelemetryForTest)
+	t.Setenv("AZD_CONFIG_DIR", t.TempDir())
+	t.Setenv(collectTelemetryEnvVar, "yes")
+
 	type args struct {
 		ctx                context.Context
 		enableDebugLogging bool
diff --git a/cli/azd/internal/tracing/fields/fields.go b/cli/azd/internal/tracing/fields/fields.go
index aacf54c77bd..86a41a8706c 100644
--- a/cli/azd/internal/tracing/fields/fields.go
+++ b/cli/azd/internal/tracing/fields/fields.go
@@ -1482,16 +1482,9 @@ var (
 		Purpose:        PerformanceAndHealth,
 		IsMeasurement:  true,
 	}
-	// CopilotMessageBillingRate is the billing rate multiplier per message.
-	CopilotMessageBillingRate = AttributeKey{
-		Key:            attribute.Key("copilot.message.billingRate"),
-		Classification: SystemMetadata,
-		Purpose:        BusinessInsight,
-		IsMeasurement:  true,
-	}
-	// CopilotMessagePremiumRequests is the number of premium requests used per message.
-	CopilotMessagePremiumRequests = AttributeKey{
-		Key:            attribute.Key("copilot.message.premiumRequests"),
+	// CopilotMessageAICredits is the number of AI credits consumed per session.
+	CopilotMessageAICredits = AttributeKey{
+		Key:            attribute.Key("copilot.message.aiCredits"),
 		Classification: SystemMetadata,
 		Purpose:        BusinessInsight,
 		IsMeasurement:  true,
diff --git a/cli/azd/pkg/azdext/contracts/v1/ai_model.pb.go b/cli/azd/pkg/azdext/contracts/v1/ai_model.pb.go
index ebc254d762c..25f76e402ff 100644
--- a/cli/azd/pkg/azdext/contracts/v1/ai_model.pb.go
+++ b/cli/azd/pkg/azdext/contracts/v1/ai_model.pb.go
@@ -28,9 +28,11 @@ type AiModel struct {
 	state  protoimpl.MessageState `protogen:"open.v1"`
 	Name   string                 `protobuf:"bytes,1,opt,name=name,proto3" json:"name,omitempty"`     // e.g. "gpt-4o"
 	Format string                 `protobuf:"bytes,2,opt,name=format,proto3" json:"format,omitempty"` // e.g. "OpenAI"
+	// Always empty. Use AiModelVersion.lifecycle_status instead.
+	//
 	// Deprecated: Marked as deprecated in azd/extensions/v1/ai_model.proto.
-	LifecycleStatus string            `protobuf:"bytes,3,opt,name=lifecycle_status,json=lifecycleStatus,proto3" json:"lifecycle_status,omitempty"` // deprecated; always empty; use AiModelVersion.lifecycle_status
-	Capabilities    []string          `protobuf:"bytes,4,rep,name=capabilities,proto3" json:"capabilities,omitempty"`                              // e.g. ["chat", "embeddings"]
+	LifecycleStatus string            `protobuf:"bytes,3,opt,name=lifecycle_status,json=lifecycleStatus,proto3" json:"lifecycle_status,omitempty"`
+	Capabilities    []string          `protobuf:"bytes,4,rep,name=capabilities,proto3" json:"capabilities,omitempty"` // e.g. ["chat", "embeddings"]
 	Versions        []*AiModelVersion `protobuf:"bytes,5,rep,name=versions,proto3" json:"versions,omitempty"`
 	Locations       []string          `protobuf:"bytes,6,rep,name=locations,proto3" json:"locations,omitempty"` // canonical locations where available
 	unknownFields   protoimpl.UnknownFields
diff --git a/cli/azd/pkg/azdext/contracts/v1beta/copilot.pb.go b/cli/azd/pkg/azdext/contracts/v1beta/copilot.pb.go
index ab0e7fc09cf..d305fba99e3 100644
--- a/cli/azd/pkg/azdext/contracts/v1beta/copilot.pb.go
+++ b/cli/azd/pkg/azdext/contracts/v1beta/copilot.pb.go
@@ -600,14 +600,21 @@ func (x *GetCopilotUsageMetricsResponse) GetUsage() *CopilotUsageMetrics {
 
 // CopilotUsageMetrics tracks resource consumption for a Copilot session.
 type CopilotUsageMetrics struct {
-	state           protoimpl.MessageState `protogen:"open.v1"`
-	Model           string                 `protobuf:"bytes,1,opt,name=model,proto3" json:"model,omitempty"`                                              // Model used.
-	InputTokens     float64                `protobuf:"fixed64,2,opt,name=input_tokens,json=inputTokens,proto3" json:"input_tokens,omitempty"`             // Total input tokens consumed.
-	OutputTokens    float64                `protobuf:"fixed64,3,opt,name=output_tokens,json=outputTokens,proto3" json:"output_tokens,omitempty"`          // Total output tokens consumed.
-	TotalTokens     float64                `protobuf:"fixed64,4,opt,name=total_tokens,json=totalTokens,proto3" json:"total_tokens,omitempty"`             // Sum of input + output tokens.
-	BillingRate     float64                `protobuf:"fixed64,5,opt,name=billing_rate,json=billingRate,proto3" json:"billing_rate,omitempty"`             // Per-request cost multiplier (e.g., 1.0x, 2.0x).
-	PremiumRequests float64                `protobuf:"fixed64,6,opt,name=premium_requests,json=premiumRequests,proto3" json:"premium_requests,omitempty"` // Number of premium requests used.
-	DurationMs      float64                `protobuf:"fixed64,7,opt,name=duration_ms,json=durationMs,proto3" json:"duration_ms,omitempty"`                // Total API duration in milliseconds.
+	state        protoimpl.MessageState `protogen:"open.v1"`
+	Model        string                 `protobuf:"bytes,1,opt,name=model,proto3" json:"model,omitempty"`                                     // Model used.
+	InputTokens  float64                `protobuf:"fixed64,2,opt,name=input_tokens,json=inputTokens,proto3" json:"input_tokens,omitempty"`    // Total input tokens consumed.
+	OutputTokens float64                `protobuf:"fixed64,3,opt,name=output_tokens,json=outputTokens,proto3" json:"output_tokens,omitempty"` // Total output tokens consumed.
+	TotalTokens  float64                `protobuf:"fixed64,4,opt,name=total_tokens,json=totalTokens,proto3" json:"total_tokens,omitempty"`    // Sum of input + output tokens.
+	// Legacy per-request cost multiplier. Use ai_credits instead.
+	//
+	// Deprecated: Marked as deprecated in azd/extensions/v1beta/copilot.proto.
+	BillingRate float64 `protobuf:"fixed64,5,opt,name=billing_rate,json=billingRate,proto3" json:"billing_rate,omitempty"`
+	// Legacy premium request count. Use ai_credits instead.
+	//
+	// Deprecated: Marked as deprecated in azd/extensions/v1beta/copilot.proto.
+	PremiumRequests float64 `protobuf:"fixed64,6,opt,name=premium_requests,json=premiumRequests,proto3" json:"premium_requests,omitempty"`
+	DurationMs      float64 `protobuf:"fixed64,7,opt,name=duration_ms,json=durationMs,proto3" json:"duration_ms,omitempty"` // Total API duration in milliseconds.
+	AiCredits       float64 `protobuf:"fixed64,8,opt,name=ai_credits,json=aiCredits,proto3" json:"ai_credits,omitempty"`    // Total AI credits consumed.
 	unknownFields   protoimpl.UnknownFields
 	sizeCache       protoimpl.SizeCache
 }
@@ -670,6 +677,7 @@ func (x *CopilotUsageMetrics) GetTotalTokens() float64 {
 	return 0
 }
 
+// Deprecated: Marked as deprecated in azd/extensions/v1beta/copilot.proto.
 func (x *CopilotUsageMetrics) GetBillingRate() float64 {
 	if x != nil {
 		return x.BillingRate
@@ -677,6 +685,7 @@ func (x *CopilotUsageMetrics) GetBillingRate() float64 {
 	return 0
 }
 
+// Deprecated: Marked as deprecated in azd/extensions/v1beta/copilot.proto.
 func (x *CopilotUsageMetrics) GetPremiumRequests() float64 {
 	if x != nil {
 		return x.PremiumRequests
@@ -691,6 +700,13 @@ func (x *CopilotUsageMetrics) GetDurationMs() float64 {
 	return 0
 }
 
+func (x *CopilotUsageMetrics) GetAiCredits() float64 {
+	if x != nil {
+		return x.AiCredits
+	}
+	return 0
+}
+
 // GetCopilotFileChangesRequest requests cached file changes for a session.
 type GetCopilotFileChangesRequest struct {
 	state         protoimpl.MessageState `protogen:"open.v1"`
@@ -1073,16 +1089,18 @@ const file_azd_extensions_v1beta_copilot_proto_rawDesc = "" +
 	"\n" +
 	"session_id\x18\x01 \x01(\tR\tsessionId\"b\n" +
 	"\x1eGetCopilotUsageMetricsResponse\x12@\n" +
-	"\x05usage\x18\x01 \x01(\v2*.azd.extensions.v1beta.CopilotUsageMetricsR\x05usage\"\x85\x02\n" +
+	"\x05usage\x18\x01 \x01(\v2*.azd.extensions.v1beta.CopilotUsageMetricsR\x05usage\"\xac\x02\n" +
 	"\x13CopilotUsageMetrics\x12\x14\n" +
 	"\x05model\x18\x01 \x01(\tR\x05model\x12!\n" +
 	"\finput_tokens\x18\x02 \x01(\x01R\vinputTokens\x12#\n" +
 	"\routput_tokens\x18\x03 \x01(\x01R\foutputTokens\x12!\n" +
-	"\ftotal_tokens\x18\x04 \x01(\x01R\vtotalTokens\x12!\n" +
-	"\fbilling_rate\x18\x05 \x01(\x01R\vbillingRate\x12)\n" +
-	"\x10premium_requests\x18\x06 \x01(\x01R\x0fpremiumRequests\x12\x1f\n" +
+	"\ftotal_tokens\x18\x04 \x01(\x01R\vtotalTokens\x12%\n" +
+	"\fbilling_rate\x18\x05 \x01(\x01B\x02\x18\x01R\vbillingRate\x12-\n" +
+	"\x10premium_requests\x18\x06 \x01(\x01B\x02\x18\x01R\x0fpremiumRequests\x12\x1f\n" +
 	"\vduration_ms\x18\a \x01(\x01R\n" +
-	"durationMs\"=\n" +
+	"durationMs\x12\x1d\n" +
+	"\n" +
+	"ai_credits\x18\b \x01(\x01R\taiCredits\"=\n" +
 	"\x1cGetCopilotFileChangesRequest\x12\x1d\n" +
 	"\n" +
 	"session_id\x18\x01 \x01(\tR\tsessionId\"l\n" +
diff --git a/docs/reference/telemetry-data.md b/docs/reference/telemetry-data.md
index 240ebb695b1..794710f7b53 100644
--- a/docs/reference/telemetry-data.md
+++ b/docs/reference/telemetry-data.md
@@ -474,8 +474,7 @@ Emitted at provision start by the `microsoft.foundry` provisioning provider (the
 | `copilot.message.model` | string | Model for specific message |
 | `copilot.message.inputTokens` | measurement | Input token count |
 | `copilot.message.outputTokens` | measurement | Output token count |
-| `copilot.message.billingRate` | measurement | Billing rate |
-| `copilot.message.premiumRequests` | measurement | Premium request count |
+| `copilot.message.aiCredits` | measurement | AI credits consumed during the session |
 | `copilot.message.durationMs` | measurement | Message duration |
 | `copilot.consent.approvedCount` | measurement | Approved consent actions |
 | `copilot.consent.deniedCount` | measurement | Denied consent actions |
diff --git a/docs/specs/metrics-audit/feature-telemetry-matrix.md b/docs/specs/metrics-audit/feature-telemetry-matrix.md
index d23449e379b..3c31b379443 100644
--- a/docs/specs/metrics-audit/feature-telemetry-matrix.md
+++ b/docs/specs/metrics-audit/feature-telemetry-matrix.md
@@ -33,7 +33,7 @@ These commands emit attributes or events beyond the global middleware span.
 | Extensions (dynamic) | `extension.id`, `extension.version`, `extension.event`, `extension.version.from`, `extension.version.to`, `extension.source.category`, `extension.source.kind`, `extension.source.category.from`, `extension.source.category.to`, `extension.installed.source.category`, `extension.dependency_of`, `extension.dependency_update_count`, `extension.update.outcome`, `extension.update.duration_ms`, `extension.grpc.legacy_call_count`, `error.chain.types`, `error.extension.cause_types`, `error.mapper.source.type`, `error.mapper.destination.type` + trace-context propagation to child process | Covers `ext.run`, `ext.install`, `ext.update`, `ext.promote`, source registration, installed inventory, failed-invocation attribution, and temporary legacy gRPC bridge use without emitting source names or locations |
 | `mcp start` | Per-tool spans via `tracing.Start` with `mcp.client.name`, `mcp.client.version` | MCP event prefix `mcp.*` |
 | `tool install` / `tool update` / `tool uninstall` / `tool check` / `tool list` / `tool show` | `tool.id`, `tool.ids`, `tool.dry_run`, `tool.install.strategy`, `tool.install.success`, `tool.install.success_count`, `tool.install.failure_count`, `tool.install.failed_ids`, `tool.install.duration_ms`, `tool.update.from_version`, `tool.update.to_version`, `tool.check.updates_available` | Comprehensive coverage in `cli/azd/cmd/tool.go`; install/update emit `tools.pack.build` spans for pack-based tools |
-| `copilot` (agent) | `copilot.initialize` event (model + reasoning config), `copilot.session` event (session create/resume) | Emitted from `internal/agent/copilot_agent.go`; covers the experimental copilot agent surface |
+| `copilot` (agent) | `copilot.initialize` event (model + reasoning config), `copilot.session` event (session create/resume), `copilot.message.aiCredits` | Emitted from `internal/agent/copilot_agent.go`; covers the experimental copilot agent surface and cumulative AI-credit usage |
 | `provision` | `validation.provision` event (provision validation outcome + 6 fields), 8 `arm.*` events (subscription / resource-group deploy / stack-deploy / what-if / validate), `aks.postprovision.skip`, per-layer `provision.layer.*` counts (`count`, `max_parallel`, `safe_fallback_count`, `explicit_dependson_count`) when multi-layer infra is used | Telemetry added across `internal/cmd/provision_*.go` and the ARM deployment client |
 | `deploy` / `publish` / `package` | `deploy.appservice.zip` event (zip-deploy outcome), `container.credentials` / `container.publish` / `container.remotebuild` events for container-based services | Per-service-target instrumentation; container events emitted from container-app and ACR push paths |
 | `hooks run` (and all hook-running commands) | `hooks.exec` event with `hooks.name` (hashed unless built-in lifecycle name), `hooks.type` (project / service / **layer**), `hooks.kind` (script runtime — `sh` / `pwsh` / `js` / `ts` / `python` / `dotnet`) | `hooks.type=layer` was added with multi-layer provision; pre/post is encoded in `hooks.name` (e.g., `prebuild` / `postbuild`); emitted from the hooks runner on every lifecycle command |
@@ -103,7 +103,7 @@ These commands emit attributes or events beyond the global middleware span.
 | `tool check` | — | ✅ | ✅ | ❌ | `tool.check.updates_available` (count) |
 | `tool show` | — | ✅ | ✅ | ❌ | `tool.id` |
 | **Copilot (Agent)** | | | | | |
-| `copilot` | — | ✅ | ✅ | ✅ | `copilot.initialize` event captures model + reasoning config; `copilot.session` event tracks session create/resume |
+| `copilot` | — | ✅ | ✅ | ✅ | `copilot.initialize` captures model + reasoning config; `copilot.session` tracks session create/resume; `copilot.message.aiCredits` measures cumulative usage |
 | **Disabled** | | | | | |
 | `version` | — | 🚫 | — | — | Intentionally disabled |
 | `telemetry upload` | — | 🚫 | — | — | Intentionally disabled |
diff --git a/docs/specs/metrics-audit/telemetry-schema.md b/docs/specs/metrics-audit/telemetry-schema.md
index 1b4f8ecbdba..c8c57bacc39 100644
--- a/docs/specs/metrics-audit/telemetry-schema.md
+++ b/docs/specs/metrics-audit/telemetry-schema.md
@@ -354,8 +354,7 @@ set on its own.
 | Model | `copilot.message.model` | SystemMetadata | FeatureInsight | |
 | Input tokens | `copilot.message.inputTokens` | SystemMetadata | PerformanceAndHealth | **Measurement** |
 | Output tokens | `copilot.message.outputTokens` | SystemMetadata | PerformanceAndHealth | **Measurement** |
-| Billing rate | `copilot.message.billingRate` | SystemMetadata | BusinessInsight | **Measurement** |
-| Premium requests | `copilot.message.premiumRequests` | SystemMetadata | BusinessInsight | **Measurement** |
+| AI credits | `copilot.message.aiCredits` | SystemMetadata | BusinessInsight | **Measurement** |
 | Duration (ms) | `copilot.message.durationMs` | SystemMetadata | PerformanceAndHealth | **Measurement** |
 
 ### Copilot Consent