diff --git a/packages/client/connection/src/client/fixture.ts b/packages/client/connection/src/client/fixture.ts
index 0843341191..bf26e96f31 100644
--- a/packages/client/connection/src/client/fixture.ts
+++ b/packages/client/connection/src/client/fixture.ts
@@ -1137,16 +1137,17 @@ export function createFixtureApi(options: FixtureOptions = {}): ApiProxy {
         }
         append(id, { type: 'user/message', surfaceOp: 'append', data: userMessage(content) })
         const target = modelTargets.get(id) ?? { provider: 'deepseek', model: 'deepseek-v4-flash' }
-        const usage = tokenUsageOf(logOf(id))
         emitMux({
           type: 'session/model-request',
           sessionId: id,
-          turn,
+          // The fixture's durable transcript is historically zero-based, while
+          // the real Agent's request telemetry opens turns at one.
+          turn: turn + 1,
           step: 0,
           provider: target.provider,
           model: target.model,
-          contextTokens: usage.uncachedInputTokens + usage.outputTokens
-            + usage.cacheReadTokens + usage.cacheWriteTokens,
+          // No fixture token-meter is composed, so omit the request-pressure
+          // numerator instead of substituting cumulative provider billing.
           contextWindow: 128_000,
         })
         startReply(
diff --git a/packages/client/connection/tests/fixture.spec.ts b/packages/client/connection/tests/fixture.spec.ts
index d25a4974ca..90d23f98cd 100644
--- a/packages/client/connection/tests/fixture.spec.ts
+++ b/packages/client/connection/tests/fixture.spec.ts
@@ -197,11 +197,10 @@ describe('createFixtureApi', () => {
     expect(frames).toContainEqual({
       type: 'session/model-request',
       sessionId: id,
-      turn: 0,
+      turn: 1,
       step: 0,
       provider: 'deepseek',
       model: 'deepseek-v4-flash',
-      contextTokens: 0,
       contextWindow: 128_000,
     })
     expect(frames.some(frame =>
diff --git a/packages/client/runtime/src/client/index.ts b/packages/client/runtime/src/client/index.ts
index 117a86ccf2..71f9f78d9f 100644
--- a/packages/client/runtime/src/client/index.ts
+++ b/packages/client/runtime/src/client/index.ts
@@ -154,11 +154,11 @@ export function apply(ctx: Context): void {
       workspaces.handleConnected()
       ctx.emit('connection/reset')
     },
-    onStateChange: (state) => {
+    onDisconnected: () => {
       // Generation death fires before any next-generation frame can arrive
       // (reconnect replays flow from stream open, ahead of onConnected):
       // the only safe moment to drop generation-scoped interaction state.
-      if (state === 'reconnecting') sessions.handleDisconnected()
+      sessions.handleDisconnected()
     },
   })
   ctx.effect(() => () => { loop.stop() }, 'runtime: connection stream loop')
diff --git a/packages/client/runtime/tests/client-apply.spec.ts b/packages/client/runtime/tests/client-apply.spec.ts
index 1e2937b464..e219a31428 100644
--- a/packages/client/runtime/tests/client-apply.spec.ts
+++ b/packages/client/runtime/tests/client-apply.spec.ts
@@ -102,7 +102,7 @@ describe('runtime client apply', () => {
     expect(bench.api.callsOf('session.create')).toHaveLength(1)
   })
 
-  it('clears connection-local request telemetry on reconnect but not connected', async () => {
+  it('clears connection-local request telemetry after every disconnected generation but not connected', async () => {
     const bench = await mount()
     const sessions = bench.ctx.get('sessions') as SessionsService
     bench.sinks?.onHostEnvelope?.({
@@ -118,7 +118,7 @@ describe('runtime client apply', () => {
         type: 'session/model-request',
         sessionId: 's-state',
         turn: 1,
-        step: 1,
+        step: 0,
         provider: 'test',
         model: 'alpha',
         contextTokens: 32_000,
@@ -133,7 +133,27 @@ describe('runtime client apply', () => {
       contextWindow: 128_000,
     })
 
-    bench.sinks?.onStateChange?.('reconnecting')
+    bench.sinks?.onDisconnected?.()
+    expect(session.getSnapshot().modelRequest).toBeNull()
+
+    // A second failed generation does not produce another deduplicated
+    // `reconnecting` state transition, but its own disconnect callback still
+    // clears telemetry received before that generation's handshake failed.
+    bench.sinks?.onMuxEnvelope?.({
+      rpcId: 'request-2' as never,
+      payload: {
+        type: 'session/model-request',
+        sessionId: 's-state',
+        turn: 2,
+        step: 0,
+        provider: 'test',
+        model: 'beta',
+        contextTokens: 48_000,
+        contextWindow: 256_000,
+      } as never,
+    })
+    expect(session.getSnapshot().modelRequest?.model).toBe('beta')
+    bench.sinks?.onDisconnected?.()
     expect(session.getSnapshot().modelRequest).toBeNull()
   })
 
diff --git a/packages/host/apiproxy/package.json b/packages/host/apiproxy/package.json
index 8cbdb91cf4..30188d66fb 100644
--- a/packages/host/apiproxy/package.json
+++ b/packages/host/apiproxy/package.json
@@ -66,6 +66,7 @@
   "devDependencies": {
     "@deepseek-ai/dsh-storage": "workspace:^",
     "@deepseek-ai/dsh-storage-domain": "workspace:^",
+    "@deepseek-ai/dsh-token-meter": "workspace:^",
     "cordis": "^4.0.0-rc.7",
     "@deepseek-ai/dsh-invariants": "workspace:^"
   }
diff --git a/packages/host/apiproxy/src/api-proxy.ts b/packages/host/apiproxy/src/api-proxy.ts
index 6658ad4e86..1367a83400 100644
--- a/packages/host/apiproxy/src/api-proxy.ts
+++ b/packages/host/apiproxy/src/api-proxy.ts
@@ -32,6 +32,8 @@ import type {
 import type {} from '@deepseek-ai/dsh-session-projection'
 // Type-only: resolves `ctx.get('sessionProjectionCache')` (the cold listing column).
 import type {} from '@deepseek-ai/dsh-session-projection-cache'
+// Type-only: resolves the optional `ctx.get('tokenMeter')` service seam.
+import type {} from '@deepseek-ai/dsh-token-meter'
 // GoalError narrows domain rejections to their stable codes at the wire boundary.
 import { GoalError } from '@deepseek-ai/dsh-goal'
 import type { GoalRef as CoreGoalRef } from '@deepseek-ai/dsh-goal'
@@ -495,34 +497,30 @@ export function createApiProxy(ctx: Context, defaults: ApiProxyDefaults): ApiPro
     for (const queue of muxQueues) queue.push(envelope)
   }
 
-  ctx.effect(() => {
-    return ctx.on('agent/model-request', (agent, turn, step, request) => {
-      const tokenMeter = ctx.get('tokenMeter') as {
-        measure(session: Session): { totalTokens: number }
-      } | undefined
-      let contextTokens: number | undefined
-      if (tokenMeter !== undefined) {
-        try {
-          contextTokens = tokenMeter.measure(agent.session).totalTokens
-        } catch {
-          // A malformed or temporarily unmeasurable replay omits only the
-          // numerator; this request still replaces stale telemetry.
-        }
+  ctx.on('agent/model-request', (agent, turn, step, request) => {
+    const tokenMeter = ctx.get('tokenMeter')
+    let contextTokens: number | undefined
+    if (tokenMeter !== undefined) {
+      try {
+        contextTokens = tokenMeter.measure(agent.session).totalTokens
+      } catch {
+        // A malformed or temporarily unmeasurable replay omits only the
+        // numerator; this request still replaces stale telemetry.
       }
-      broadcast({
-        type: 'session/model-request',
-        sessionId: agent.session.id,
-        turn,
-        step,
-        provider: request.provider,
-        model: request.model,
-        ...contextTokens === undefined ? {} : { contextTokens },
-        ...request.contextWindow === undefined
-          ? {}
-          : { contextWindow: request.contextWindow },
-      })
+    }
+    broadcast({
+      type: 'session/model-request',
+      sessionId: agent.session.id,
+      turn,
+      step,
+      provider: request.provider,
+      model: request.model,
+      ...contextTokens === undefined ? {} : { contextTokens },
+      ...request.contextWindow === undefined
+        ? {}
+        : { contextWindow: request.contextWindow },
     })
-  }, 'api-proxy: model request telemetry')
+  })
 
   // Projection change feed → session/projection push frames. The carrier
   // mints the wire frame (the seam package holds no wire vocabulary); the
diff --git a/packages/host/apiproxy/src/api/events.schema.ts b/packages/host/apiproxy/src/api/events.schema.ts
index b7dd5121e8..e94abf6ca2 100644
--- a/packages/host/apiproxy/src/api/events.schema.ts
+++ b/packages/host/apiproxy/src/api/events.schema.ts
@@ -41,7 +41,7 @@ export const muxFrameSchema = z.discriminatedUnion('type', [
     type: z.literal('session/model-request'),
     sessionId: sessionIdSchema,
     turn: z.number().int().positive(),
-    step: z.number().int().positive(),
+    step: z.number().int().nonnegative(),
     provider: z.string().min(1),
     model: z.string().min(1),
     contextTokens: z.number().int().nonnegative().optional(),
diff --git a/packages/host/apiproxy/tests/api-proxy-model-request.spec.ts b/packages/host/apiproxy/tests/api-proxy-model-request.spec.ts
index 713e9b1fa4..0fc44ba1ed 100644
--- a/packages/host/apiproxy/tests/api-proxy-model-request.spec.ts
+++ b/packages/host/apiproxy/tests/api-proxy-model-request.spec.ts
@@ -36,7 +36,7 @@ describe('ApiProxy model-request telemetry', () => {
     } as Agent
     ctx.agents.register(agent)
     const measure = vi.fn(() => ({ totalTokens: 321 }))
-    const removeTokenMeter = ctx.provide('tokenMeter' as never, { measure } as never)
+    const removeTokenMeter = ctx.provide('tokenMeter', { measure })
     const api = createApiProxy(ctx, {
       provider: 'test',
       model: 'alpha',
diff --git a/packages/host/apiproxy/tests/rpc-schemas.spec.ts b/packages/host/apiproxy/tests/rpc-schemas.spec.ts
index 88f8f4c24a..7046780b48 100644
--- a/packages/host/apiproxy/tests/rpc-schemas.spec.ts
+++ b/packages/host/apiproxy/tests/rpc-schemas.spec.ts
@@ -365,7 +365,7 @@ describe('events frame schemas', () => {
         type: 'session/model-request',
         sessionId: 's',
         turn: 2,
-        step: 1,
+        step: 0,
         provider: 'deepseek',
         model: 'deepseek-chat',
         contextTokens: 8_000,
@@ -392,6 +392,7 @@ describe('events frame schemas', () => {
     expect(() => muxFrameSchema.parse({ type: 'unknown/frame' })).toThrow()
     for (const invalid of [
       { type: 'session/model-request', sessionId: 's', turn: 0, step: 1, provider: 'p', model: 'm' },
+      { type: 'session/model-request', sessionId: 's', turn: 1, step: -1, provider: 'p', model: 'm' },
       { type: 'session/model-request', sessionId: 's', turn: 1, step: 1, provider: 'p', model: 'm', contextTokens: -1 },
       { type: 'session/model-request', sessionId: 's', turn: 1, step: 1, provider: 'p', model: 'm', contextWindow: 0 },
       { type: 'session/projection', sessionId: 's', key: '', value: null, seq: 0 },
diff --git a/packages/host/apiproxy/tsconfig.json b/packages/host/apiproxy/tsconfig.json
index 7e1e83e39b..19483b6bc0 100644
--- a/packages/host/apiproxy/tsconfig.json
+++ b/packages/host/apiproxy/tsconfig.json
@@ -23,6 +23,9 @@
     {
       "path": "../../llm/llm"
     },
+    {
+      "path": "../../llm/token-meter"
+    },
     {
       "path": "../../core/agent"
     },
diff --git a/packages/llm/token-meter/src/invariant.ts b/packages/llm/token-meter/src/invariant.ts
index b8f0cc385a..e0bbe268e1 100644
--- a/packages/llm/token-meter/src/invariant.ts
+++ b/packages/llm/token-meter/src/invariant.ts
@@ -15,8 +15,11 @@ export const name = 'token-meter-invariant'
 export const inject = ['invariants']
 
 /**
- * No runtime invariant: token estimates are per-call outputs and the private session cache is
- * invalidated at its event mutation boundary; neither exposes an independent observation stream.
+ * No runtime invariant: token estimates are per-call outputs and the private
+ * session cache is invalidated at its event mutation boundary. The package's
+ * projection does expose an observation stream, but its schema fixes the JSON
+ * payload and its pure fold replaces same-step samples; totals need not be
+ * monotone when a final usage sample corrects an earlier chunk.
  */
 const install: InvariantInstaller = () => {}
 
diff --git a/pnpm-lock.yaml b/pnpm-lock.yaml
index 1fe5595d18..93469a1947 100644
--- a/pnpm-lock.yaml
+++ b/pnpm-lock.yaml
@@ -2924,6 +2924,9 @@ importers:
       '@deepseek-ai/dsh-storage-domain':
         specifier: workspace:^
         version: link:../../storage/storage-domain
+      '@deepseek-ai/dsh-token-meter':
+        specifier: workspace:^
+        version: link:../../llm/token-meter
       cordis:
         specifier: ^4.0.0-rc.7
         version: 4.0.0-rc.7(@cordisjs/plugin-include@1.0.4)(@cordisjs/plugin-loader@1.0.0-rc.5)