diff --git a/api/server/controllers/agents/__tests__/usageEvents.live.spec.js b/api/server/controllers/agents/__tests__/usageEvents.live.spec.js
new file mode 100644
index 0000000000..ff02c9325f
--- /dev/null
+++ b/api/server/controllers/agents/__tests__/usageEvents.live.spec.js
@@ -0,0 +1,161 @@
+/**
+ * Live host-layer verification: real Anthropic run through the actual
+ * getDefaultHandlers pipeline, asserting the SSE usage/context events the
+ * client consumes and their resume persistence.
+ *
+ * Run with:
+ * RUN_USAGE_LIVE_TESTS=1 ANTHROPIC_API_KEY=... npx jest usageEvents.live --runInBand
+ */
+const { HumanMessage } = require('@langchain/core/messages');
+const { Run, Providers, GraphEvents } = require('@librechat/agents');
+const { GenerationJobManager } = require('@librechat/api');
+const { getDefaultHandlers } = require('~/server/controllers/agents/callbacks');
+
+jest.mock('nanoid', () => ({
+  nanoid: jest.fn(() => 'mock-nanoid'),
+}));
+
+jest.mock('~/server/services/Files/Citations', () => ({
+  processFileCitations: jest.fn(),
+}));
+
+jest.mock('~/server/services/Files/Code/process', () => ({
+  processCodeOutput: jest.fn(),
+  runPreviewFinalize: jest.fn(),
+}));
+
+jest.mock('~/server/services/Files/process', () => ({
+  saveBase64Image: jest.fn(),
+}));
+
+const shouldRunLive =
+  process.env.RUN_USAGE_LIVE_TESTS === '1' &&
+  process.env.ANTHROPIC_API_KEY != null &&
+  process.env.ANTHROPIC_API_KEY !== '';
+
+const describeIfLive = shouldRunLive ? describe : describe.skip;
+const modelName = process.env.ANTHROPIC_USAGE_LIVE_MODEL ?? 'claude-haiku-4-5';
+const hasContextUsageEvent = GraphEvents.ON_CONTEXT_USAGE != null;
+
+const charCounter = (msg) => {
+  const content = msg.content;
+  if (typeof content === 'string') {
+    return Math.ceil(content.length / 4) + 3;
+  }
+  if (Array.isArray(content)) {
+    let length = 3;
+    for (const part of content) {
+      if (typeof part === 'string') {
+        length += Math.ceil(part.length / 4);
+      } else if (typeof part?.text === 'string') {
+        length += Math.ceil(part.text.length / 4);
+      }
+    }
+    return length;
+  }
+  return 3;
+};
+
+function createMockRes() {
+  const events = [];
+  return {
+    events,
+    headersSent: true,
+    writableEnded: false,
+    write(payload) {
+      for (const line of String(payload).split('\n')) {
+        if (line.startsWith('data: ')) {
+          events.push(JSON.parse(line.slice(6)));
+        }
+      }
+      return true;
+    },
+  };
+}
+
+describeIfLive('live usage events through the host pipeline', () => {
+  jest.setTimeout(120000);
+
+  afterAll(async () => {
+    await GenerationJobManager.destroy();
+  });
+
+  test('streams real provider usage and persists it for resume', async () => {
+    const streamId = `usage-live-${Date.now()}`;
+    await GenerationJobManager.createJob(streamId, 'user-live', 'convo-live');
+
+    /** streamId mode routes events through the job emitter — capture them
+     *  as a subscribed resumable client would, not via res.write */
+    const res = createMockRes();
+    await GenerationJobManager.subscribe(streamId, (event) => {
+      res.events.push(event);
+    });
+    const collectedUsage = [];
+    const handlers = getDefaultHandlers({
+      res,
+      aggregateContent: () => {},
+      toolEndCallback: () => {},
+      collectedUsage,
+      streamId,
+    });
+
+    const run = await Run.create({
+      runId: 'usage-live-response',
+      graphConfig: {
+        type: 'standard',
+        llmConfig: {
+          provider: Providers.ANTHROPIC,
+          model: modelName,
+          apiKey: process.env.ANTHROPIC_API_KEY,
+          temperature: 0,
+          maxTokens: 64,
+          streaming: true,
+          streamUsage: true,
+        },
+        instructions: 'You are concise. Reply with one short sentence.',
+        maxContextTokens: 8000,
+      },
+      returnContent: true,
+      customHandlers: handlers,
+      tokenCounter: charCounter,
+      indexTokenCountMap: {},
+    });
+
+    await run.processStream(
+      { messages: [new HumanMessage('Say hello in five words or fewer.')] },
+      {
+        configurable: { thread_id: 'usage-live-thread', user_id: 'user-live' },
+        streamMode: 'values',
+        version: 'v2',
+      },
+    );
+
+    const usageEvents = res.events.filter((e) => e.event === 'on_token_usage');
+    expect(usageEvents).toHaveLength(1);
+    const usage = usageEvents[0].data;
+    expect(usage.input_tokens).toBeGreaterThan(0);
+    expect(usage.output_tokens).toBeGreaterThan(0);
+    expect(usage.provider).toBe(Providers.ANTHROPIC);
+    expect(usage.model).toBe(modelName);
+    expect(collectedUsage).toHaveLength(1);
+    expect(usage.input_tokens).toBe(collectedUsage[0].input_tokens);
+
+    if (hasContextUsageEvent) {
+      const contextEvents = res.events.filter((e) => e.event === 'on_context_usage');
+      expect(contextEvents).toHaveLength(1);
+      const snapshot = contextEvents[0].data;
+      expect(snapshot.breakdown.maxContextTokens).toBe(8000);
+      const estimatedUsed = snapshot.contextBudget - snapshot.remainingContextTokens;
+      expect(estimatedUsed).toBeGreaterThan(0);
+      expect(estimatedUsed / usage.input_tokens).toBeGreaterThan(0.2);
+      expect(estimatedUsed / usage.input_tokens).toBeLessThan(5);
+    }
+
+    const resumeState = await GenerationJobManager.getResumeState(streamId);
+    expect(resumeState.collectedUsage).toHaveLength(1);
+    expect(resumeState.collectedUsage[0].input_tokens).toBe(usage.input_tokens);
+    if (hasContextUsageEvent) {
+      expect(resumeState.contextUsage.breakdown.maxContextTokens).toBe(8000);
+    }
+  });
+});