chore(main): release 0.25.2 (#511)

🤖 I have created a release *beep* *boop*
---


##
[0.25.2](https://github.com/agentclientprotocol/claude-agent-acp/compare/v0.25.1...v0.25.2)
(2026-04-06)


### Bug Fixes

* prioritize ANTHROPIC_MODEL env var over settings.model in model …
([#505](https://github.com/agentclientprotocol/claude-agent-acp/issues/505))
([bea1a40](bea1a40e9b))

---
This PR was generated with [Release
Please](https://github.com/googleapis/release-please). See
[documentation](https://github.com/googleapis/release-please#release-please).

Co-authored-by: acp-release-bot[bot] <246668977+acp-release-bot[bot]@users.noreply.github.com>

.release-please-manifest.json

@@ -1,3 +1,3 @@
 {
-  ".": "0.24.2"
+  ".": "0.25.2"
 }

CHANGELOG.md

@@ -1,5 +1,35 @@
 # Changelog
 
+## [0.25.2](https://github.com/agentclientprotocol/claude-agent-acp/compare/v0.25.1...v0.25.2) (2026-04-06)
+
+
+### Bug Fixes
+
+* prioritize ANTHROPIC_MODEL env var over settings.model in model … ([#505](https://github.com/agentclientprotocol/claude-agent-acp/issues/505)) ([bea1a40](https://github.com/agentclientprotocol/claude-agent-acp/commit/bea1a40e9bfe1e06672b118a727f9339def3be23))
+
+## [0.25.1](https://github.com/agentclientprotocol/claude-agent-acp/compare/v0.25.0...v0.25.1) (2026-04-06)
+
+
+### Bug Fixes
+
+* add `auto` to valid modes in applySessionMode to fix mode cycling ([#507](https://github.com/agentclientprotocol/claude-agent-acp/issues/507)) ([15e91fb](https://github.com/agentclientprotocol/claude-agent-acp/commit/15e91fb5c3449de2600b583cb7a8d36b5b510443))
+
+## [0.25.0](https://github.com/agentclientprotocol/claude-agent-acp/compare/v0.24.2...v0.25.0) (2026-04-03)
+
+
+### Features
+
+* Add auto permission mode support ([#501](https://github.com/agentclientprotocol/claude-agent-acp/issues/501)) ([a161453](https://github.com/agentclientprotocol/claude-agent-acp/commit/a16145396fe2e6ead8734478961b2de65707e335))
+* Add separate Claude and Console terminal logins ([#502](https://github.com/agentclientprotocol/claude-agent-acp/issues/502)) ([063cd35](https://github.com/agentclientprotocol/claude-agent-acp/commit/063cd353809df8f3dda9e57d8ea848c0a6d44257))
+* Update to claude-agent-sdk 0.2.91 ([#500](https://github.com/agentclientprotocol/claude-agent-acp/issues/500)) ([65a2230](https://github.com/agentclientprotocol/claude-agent-acp/commit/65a223038576d72b74e1483fed10e982a1f842bd))
+
+
+### Bug Fixes
+
+* log warnings for malformed settings files instead of silent fallback ([#486](https://github.com/agentclientprotocol/claude-agent-acp/issues/486)) ([ae6c388](https://github.com/agentclientprotocol/claude-agent-acp/commit/ae6c38831415f9fc1de2d3dd1d4a247becbbd32f))
+* prevent race conditions in SettingsManager setCwd and debounce ([#485](https://github.com/agentclientprotocol/claude-agent-acp/issues/485)) ([7506223](https://github.com/agentclientprotocol/claude-agent-acp/commit/7506223cffb1aba4b4560feda11f69a1395a8c9d))
+* use current model's context window for usage_update size ([#412](https://github.com/agentclientprotocol/claude-agent-acp/issues/412)) ([d07799d](https://github.com/agentclientprotocol/claude-agent-acp/commit/d07799d7b3b4e438c8b158266c79723a0b592c07))
+
 ## [0.24.2](https://github.com/agentclientprotocol/claude-agent-acp/compare/v0.24.1...v0.24.2) (2026-03-27)
 
 

package.json

@@ -3,7 +3,7 @@
   "publishConfig": {
     "access": "public"
   },
-  "version": "0.24.2",
+  "version": "0.25.2",
   "description": "An ACP-compatible coding agent powered by the Claude Agent SDK (TypeScript)",
   "main": "dist/lib.js",
   "types": "dist/lib.d.ts",
@@ -60,22 +60,22 @@
   "author": "Zed Industries",
   "license": "Apache-2.0",
   "dependencies": {
-    "@agentclientprotocol/sdk": "0.17.0",
-    "@anthropic-ai/claude-agent-sdk": "0.2.84",
+    "@agentclientprotocol/sdk": "0.18.0",
+    "@anthropic-ai/claude-agent-sdk": "0.2.92",
     "zod": "^3.25.0 || ^4.0.0"
   },
   "devDependencies": {
-    "@anthropic-ai/sdk": "0.80.0",
+    "@anthropic-ai/sdk": "0.82.0",
     "@eslint/js": "10.0.1",
-    "@types/node": "25.5.0",
-    "@typescript-eslint/eslint-plugin": "8.57.2",
-    "@typescript-eslint/parser": "8.57.2",
-    "eslint": "10.1.0",
+    "@types/node": "25.5.2",
+    "@typescript-eslint/eslint-plugin": "8.58.0",
+    "@typescript-eslint/parser": "8.58.0",
+    "eslint": "10.2.0",
     "eslint-config-prettier": "10.1.8",
     "globals": "17.4.0",
     "prettier": "3.8.1",
     "ts-node": "10.9.2",
-    "typescript": "5.9.3",
-    "vitest": "4.1.1"
+    "typescript": "6.0.2",
+    "vitest": "4.1.2"
   }
 }

src/acp-agent.ts

@@ -48,6 +48,7 @@ import {
   listSessions,
   McpServerConfig,
   ModelInfo,
+  ModelUsage,
   Options,
   PermissionMode,
   Query,
@@ -233,6 +234,7 @@ const ALLOW_BYPASS = !IS_ROOT || !!process.env.IS_SANDBOX;
 const LOCAL_ONLY_COMMANDS = new Set(["/context", "/heapdump", "/extra-usage"]);
 
 const PERMISSION_MODE_ALIASES: Record<string, PermissionMode> = {
+  auto: "auto",
   default: "default",
   acceptedits: "acceptEdits",
   dontask: "dontAsk",
@@ -306,25 +308,42 @@ export class ClaudeAcpAgent implements Agent {
       },
     };
 
-    const terminalAuthMethod: any = {
-      description: "Run `claude /login` in the terminal",
-      name: "Log in with Claude",
-      id: "claude-login",
-      type: "terminal",
-      args: ["--cli"],
-    };
     const supportsTerminalAuth = request.clientCapabilities?.auth?.terminal === true;
+    const supportsMetaTerminalAuth = request.clientCapabilities?._meta?.["terminal-auth"] === true;
+
+    const claudeLoginMethod: any = {
+      description: "Use Claude subscription ",
+      name: "Claude Subscription",
+      id: "claude-ai-login",
+      type: "terminal",
+      args: ["--cli", "auth", "login", "--claudeai"],
+    };
+
+    const consoleLoginMethod: any = {
+      description: "Use Anthropic Console (API usage billing)",
+      name: "Anthropic Console",
+      id: "console-login",
+      type: "terminal",
+      args: ["--cli", "auth", "login", "--console"],
+    };
 
     // If client supports terminal-auth capability, use that instead.
-    const supportsMetaTerminalAuth = request.clientCapabilities?._meta?.["terminal-auth"] === true;
     if (supportsMetaTerminalAuth) {
-      terminalAuthMethod._meta = {
+      const baseArgs = process.argv.slice(1);
+      claudeLoginMethod._meta = {
         "terminal-auth": {
           command: process.execPath,
-          args: [...process.argv.slice(1), "--cli"],
+          args: [...baseArgs, "--cli", "auth", "login", "--claudeai"],
           label: "Claude Login",
         },
       };
+      consoleLoginMethod._meta = {
+        "terminal-auth": {
+          command: process.execPath,
+          args: [...baseArgs, "--cli", "auth", "login", "--console"],
+          label: "Anthropic Console Login",
+        },
+      };
     }
 
     return {
@@ -358,8 +377,9 @@ export class ClaudeAcpAgent implements Agent {
       },
       authMethods: [
         ...(!shouldHideClaudeAuth() && (supportsTerminalAuth || supportsMetaTerminalAuth)
-          ? [terminalAuthMethod]
+          ? [claudeLoginMethod]
           : []),
+        ...(supportsTerminalAuth || supportsMetaTerminalAuth ? [consoleLoginMethod] : []),
         ...(supportsGatewayAuth ? [gatewayAuthMethod] : []),
       ],
     };
@@ -468,6 +488,8 @@ export class ClaudeAcpAgent implements Agent {
     };
 
     let lastAssistantTotalUsage: number | null = null;
+    let lastAssistantModel: string | null = null;
+    let lastContextWindowSize: number = 200000;
 
     const userMessage = promptToClaude(params);
 
@@ -527,9 +549,26 @@ export class ClaudeAcpAgent implements Agent {
                 break;
               }
               case "compact_boundary": {
-                // We don't know the exact size, but since we compacted,
-                // we set it to zero. The client gets the exact size on the next message.
+                // Send used:0 immediately so the client doesn't keep showing
+                // the stale pre-compaction context size until the next turn.
+                //
+                // This is a deliberate approximation: we don't know the exact
+                // post-compaction token count (only the SDK's next API call
+                // reveals that). But used:0 is directionally correct — context
+                // just dropped dramatically — and the real value replaces it
+                // within seconds when the next result message arrives.
+                // The alternative (no update) leaves the client showing e.g.
+                // "944k/1m" right after the user sees "Compacting completed",
+                // which is confusing and wrong.
                 lastAssistantTotalUsage = 0;
+                await this.client.sessionUpdate({
+                  sessionId: message.session_id,
+                  update: {
+                    sessionUpdate: "usage_update",
+                    used: 0,
+                    size: lastContextWindowSize,
+                  },
+                });
                 await this.client.sessionUpdate({
                   sessionId: message.session_id,
                   update: {
@@ -578,10 +617,11 @@ export class ClaudeAcpAgent implements Agent {
             session.accumulatedUsage.cachedReadTokens += message.usage.cache_read_input_tokens;
             session.accumulatedUsage.cachedWriteTokens += message.usage.cache_creation_input_tokens;
 
-            // Calculate context window size from modelUsage (minimum across all models used)
-            const contextWindows = Object.values(message.modelUsage).map((m) => m.contextWindow);
-            const contextWindowSize =
-              contextWindows.length > 0 ? Math.min(...contextWindows) : 200000;
+            const matchingModelUsage = lastAssistantModel
+              ? getMatchingModelUsage(message.modelUsage, lastAssistantModel)
+              : null;
+            const contextWindowSize = matchingModelUsage?.contextWindow ?? 200000;
+            lastContextWindowSize = contextWindowSize;
 
             // Send usage_update notification
             if (lastAssistantTotalUsage !== null) {
@@ -707,6 +747,11 @@ export class ClaudeAcpAgent implements Agent {
             }
 
             // Store latest assistant usage (excluding subagents)
+            // Sum all token types as a proxy for post-turn context occupancy:
+            // current turn's output will become next turn's input.
+            // Note: per the Anthropic API, input_tokens excludes cache tokens —
+            // cache_read and cache_creation are reported separately, so summing
+            // all four fields is not double-counting.
             if ((message.message as any).usage && message.parent_tool_use_id === null) {
               const messageWithUsage = message.message as unknown as SDKResultMessage;
               lastAssistantTotalUsage =
@@ -715,6 +760,16 @@ export class ClaudeAcpAgent implements Agent {
                 messageWithUsage.usage.cache_read_input_tokens +
                 messageWithUsage.usage.cache_creation_input_tokens;
             }
+            // Track the current top-level model for context window size lookup
+            // (exclude subagent messages to stay in sync with lastAssistantTotalUsage)
+            if (
+              message.type === "assistant" &&
+              message.parent_tool_use_id === null &&
+              message.message.model &&
+              message.message.model !== "<synthetic>"
+            ) {
+              lastAssistantModel = message.message.model;
+            }
 
             // Slash commands like /compact can generate invalid output... doesn't match
             // their own docs: https://docs.anthropic.com/en/docs/claude-code/sdk/sdk-slash-commands#%2Fcompact-compact-conversation-history
@@ -951,6 +1006,7 @@ export class ClaudeAcpAgent implements Agent {
 
   private async applySessionMode(sessionId: string, modeId: string): Promise<void> {
     switch (modeId) {
+      case "auto":
       case "default":
       case "acceptEdits":
       case "bypassPermissions":
@@ -1428,6 +1484,11 @@ export class ClaudeAcpAgent implements Agent {
     const models = await getAvailableModels(q, initializationResult.models, settingsManager);
 
     const availableModes = [
+      {
+        id: "auto",
+        name: "Auto",
+        decription: "Use a model classifier to approve/deny permission prompts.",
+      },
       {
         id: "default",
         name: "Default",
@@ -1638,7 +1699,16 @@ async function getAvailableModels(
 
   let currentModel = models[0];
 
-  if (settings.model) {
+  // Model priority (highest to lowest):
+  // 1. ANTHROPIC_MODEL environment variable
+  // 2. settings.model (user configuration)
+  // 3. models[0] (default first model)
+  if (process.env.ANTHROPIC_MODEL) {
+    const match = resolveModelPreference(models, process.env.ANTHROPIC_MODEL);
+    if (match) {
+      currentModel = match;
+    }
+  } else if (settings.model) {
     const match = resolveModelPreference(models, settings.model);
     if (match) {
       currentModel = match;
@@ -2109,3 +2179,28 @@ export function runAcp() {
   const stream = ndJsonStream(input, output);
   new AgentSideConnection((client) => new ClaudeAcpAgent(client), stream);
 }
+
+function commonPrefixLength(a: string, b: string) {
+  let i = 0;
+  while (i < a.length && i < b.length && a[i] === b[i]) {
+    i++;
+  }
+  return i;
+}
+
+function getMatchingModelUsage(modelUsage: Record<string, ModelUsage>, currentModel: string) {
+  let bestKey: string | null = null;
+  let bestLen = 0;
+
+  for (const key of Object.keys(modelUsage)) {
+    const len = commonPrefixLength(key, currentModel);
+    if (len > bestLen) {
+      bestLen = len;
+      bestKey = key;
+    }
+  }
+
+  if (bestKey) {
+    return modelUsage[bestKey];
+  }
+}

src/settings.ts

@@ -25,9 +25,14 @@ export interface ClaudeCodeSettings {
 }
 
 /**
- * Reads and parses a JSON settings file, returning an empty object if not found or invalid
+ * Reads and parses a JSON settings file, returning an empty object if not found or invalid.
+ * Silently ignores missing files (ENOENT) but logs warnings for other errors
+ * (malformed JSON, permission errors, etc.) to aid debugging.
  */
-async function loadSettingsFile(filePath: string | null): Promise<ClaudeCodeSettings> {
+async function loadSettingsFile(
+  filePath: string | null,
+  logger?: { error: (...args: any[]) => void },
+): Promise<ClaudeCodeSettings> {
   if (!filePath) {
     return {};
   }
@@ -35,7 +40,11 @@ async function loadSettingsFile(filePath: string | null): Promise<ClaudeCodeSett
   try {
     const content = await fs.promises.readFile(filePath, "utf-8");
     return JSON.parse(content) as ClaudeCodeSettings;
-  } catch {
+  } catch (error) {
+    if (error instanceof Error && "code" in error && error.code === "ENOENT") {
+      return {};
+    }
+    logger?.error(`Failed to load settings from ${filePath}:`, error);
     return {};
   }
 }
@@ -83,7 +92,9 @@ export class SettingsManager {
   private onChange?: () => void;
   private logger: { log: (...args: any[]) => void; error: (...args: any[]) => void };
   private initialized = false;
+  private disposed = false;
   private debounceTimer: ReturnType<typeof setTimeout> | null = null;
+  private initPromise: Promise<void> | null = null;
 
   constructor(cwd: string, options?: SettingsManagerOptions) {
     this.cwd = cwd;
@@ -98,10 +109,19 @@ export class SettingsManager {
     if (this.initialized) {
       return;
     }
+    if (this.initPromise) {
+      return this.initPromise;
+    }
 
-    await this.loadAllSettings();
-    this.setupWatchers();
-    this.initialized = true;
+    this.disposed = false;
+    this.initPromise = this.loadAllSettings().then(() => {
+      if (!this.disposed) {
+        this.setupWatchers();
+        this.initialized = true;
+      }
+      this.initPromise = null;
+    });
+    return this.initPromise;
   }
 
   /**
@@ -130,10 +150,10 @@ export class SettingsManager {
    */
   private async loadAllSettings(): Promise<void> {
     const [userSettings, projectSettings, localSettings, enterpriseSettings] = await Promise.all([
-      loadSettingsFile(this.getUserSettingsPath()),
-      loadSettingsFile(this.getProjectSettingsPath()),
-      loadSettingsFile(this.getLocalSettingsPath()),
-      loadSettingsFile(getManagedSettingsPath()),
+      loadSettingsFile(this.getUserSettingsPath(), this.logger),
+      loadSettingsFile(this.getProjectSettingsPath(), this.logger),
+      loadSettingsFile(this.getLocalSettingsPath(), this.logger),
+      loadSettingsFile(getManagedSettingsPath(), this.logger),
     ]);
 
     this.userSettings = userSettings;
@@ -226,9 +246,14 @@ export class SettingsManager {
 
     this.debounceTimer = setTimeout(async () => {
       this.debounceTimer = null;
+      if (this.disposed) {
+        return;
+      }
       try {
         await this.loadAllSettings();
-        this.onChange?.();
+        if (!this.disposed) {
+          this.onChange?.();
+        }
       } catch (error) {
         this.logger.error("Failed to reload settings:", error);
       }
@@ -259,7 +284,6 @@ export class SettingsManager {
 
     this.dispose();
     this.cwd = cwd;
-    this.initialized = false;
     await this.initialize();
   }
 
@@ -267,6 +291,10 @@ export class SettingsManager {
    * Disposes of file watchers and cleans up resources
    */
   dispose(): void {
+    this.disposed = true;
+    this.initialized = false;
+    this.initPromise = null;
+
     if (this.debounceTimer) {
       clearTimeout(this.debounceTimer);
       this.debounceTimer = null;
@@ -276,6 +304,5 @@ export class SettingsManager {
       watcher.close();
     }
     this.watchers = [];
-    this.initialized = false;
   }
 }

src/tests/acp-agent.test.ts

@@ -1608,3 +1608,446 @@ describe("session/close", () => {
     expect(agent.sessions["session-b"]).toBeDefined();
   });
 });
+
+describe("usage_update computation", () => {
+  function createAssistantMessage(overrides: {
+    model: string;
+    usage?: {
+      input_tokens: number;
+      output_tokens: number;
+      cache_read_input_tokens: number;
+      cache_creation_input_tokens: number;
+    };
+  }) {
+    return {
+      type: "assistant" as const,
+      parent_tool_use_id: null,
+      uuid: randomUUID(),
+      session_id: "test-session",
+      message: {
+        model: overrides.model,
+        content: [{ type: "text", text: "hello" }],
+        usage: overrides.usage ?? {
+          input_tokens: 100,
+          output_tokens: 50,
+          cache_read_input_tokens: 20,
+          cache_creation_input_tokens: 10,
+        },
+      },
+    };
+  }
+
+  function createResultMessageWithModel(overrides: {
+    modelUsage: Record<
+      string,
+      {
+        inputTokens: number;
+        outputTokens: number;
+        cacheReadInputTokens: number;
+        cacheCreationInputTokens: number;
+        webSearchRequests: number;
+        costUSD: number;
+        contextWindow: number;
+        maxOutputTokens: number;
+      }
+    >;
+  }) {
+    return {
+      type: "result" as const,
+      subtype: "success" as const,
+      stop_reason: "end_turn",
+      is_error: false,
+      result: "",
+      errors: [],
+      duration_ms: 0,
+      duration_api_ms: 0,
+      num_turns: 1,
+      total_cost_usd: 0.01,
+      usage: {
+        input_tokens: 10,
+        output_tokens: 5,
+        cache_read_input_tokens: 0,
+        cache_creation_input_tokens: 0,
+      },
+      modelUsage: overrides.modelUsage,
+      permission_denials: [],
+      uuid: randomUUID(),
+      session_id: "test-session",
+    };
+  }
+
+  function createMockAgentWithCapture() {
+    const updates: any[] = [];
+    const mockClient = {
+      sessionUpdate: async (notification: any) => {
+        updates.push(notification);
+      },
+    } as unknown as AgentSideConnection;
+    const agent = new ClaudeAcpAgent(mockClient, { log: () => {}, error: () => {} });
+    return { agent, updates };
+  }
+
+  function injectSession(agent: ClaudeAcpAgent, messages: any[]) {
+    const input = new Pushable<any>();
+    async function* messageGenerator() {
+      // Wait for the prompt to push its user message so we can replay it
+      const iter = input[Symbol.asyncIterator]();
+      const { value: userMessage, done } = await iter.next();
+      if (!done && userMessage) {
+        yield {
+          type: "user",
+          message: userMessage.message,
+          parent_tool_use_id: null,
+          uuid: userMessage.uuid,
+          session_id: "test-session",
+          isReplay: true,
+        };
+      }
+      yield* messages;
+    }
+    agent.sessions["test-session"] = {
+      query: messageGenerator() as any,
+      input,
+      cancelled: false,
+      cwd: "/test",
+      modes: {
+        currentModeId: "default",
+        availableModes: [],
+      },
+      models: {
+        currentModelId: "default",
+        availableModels: [],
+      },
+      settingsManager: {} as any,
+      accumulatedUsage: {
+        inputTokens: 0,
+        outputTokens: 0,
+        cachedReadTokens: 0,
+        cachedWriteTokens: 0,
+      },
+      configOptions: [],
+      promptRunning: false,
+      pendingMessages: new Map(),
+      nextPendingOrder: 0,
+      abortController: new AbortController(),
+    };
+  }
+
+  it("used sums all token types as post-turn context occupancy proxy", async () => {
+    const { agent, updates } = createMockAgentWithCapture();
+    injectSession(agent, [
+      createAssistantMessage({
+        model: "claude-opus-4-20250514",
+        usage: {
+          input_tokens: 1000,
+          output_tokens: 500,
+          cache_read_input_tokens: 200,
+          cache_creation_input_tokens: 100,
+        },
+      }),
+      createResultMessageWithModel({
+        modelUsage: {
+          "claude-opus-4-20250514": {
+            inputTokens: 1000,
+            outputTokens: 500,
+            cacheReadInputTokens: 200,
+            cacheCreationInputTokens: 100,
+            webSearchRequests: 0,
+            costUSD: 0.01,
+            contextWindow: 1000000,
+            maxOutputTokens: 16384,
+          },
+        },
+      }),
+      { type: "system", subtype: "session_state_changed", state: "idle" },
+    ]);
+
+    await agent.prompt({ sessionId: "test-session", prompt: [{ type: "text", text: "test" }] });
+
+    const usageUpdate = updates.find((u: any) => u.update?.sessionUpdate === "usage_update");
+    expect(usageUpdate).toBeDefined();
+    // used = input(1000) + output(500) + cache_read(200) + cache_creation(100) = 1800
+    expect(usageUpdate.update.used).toBe(1800);
+  });
+
+  it("size reflects the current model's context window, not min across all", async () => {
+    const { agent, updates } = createMockAgentWithCapture();
+    injectSession(agent, [
+      createAssistantMessage({ model: "claude-opus-4-20250514" }),
+      createResultMessageWithModel({
+        modelUsage: {
+          "claude-opus-4-20250514": {
+            inputTokens: 100,
+            outputTokens: 50,
+            cacheReadInputTokens: 20,
+            cacheCreationInputTokens: 10,
+            webSearchRequests: 0,
+            costUSD: 0.01,
+            contextWindow: 1000000,
+            maxOutputTokens: 16384,
+          },
+          "claude-sonnet-4-20250514": {
+            inputTokens: 50,
+            outputTokens: 25,
+            cacheReadInputTokens: 10,
+            cacheCreationInputTokens: 5,
+            webSearchRequests: 0,
+            costUSD: 0.005,
+            contextWindow: 200000,
+            maxOutputTokens: 16384,
+          },
+        },
+      }),
+      { type: "system", subtype: "session_state_changed", state: "idle" },
+    ]);
+
+    await agent.prompt({ sessionId: "test-session", prompt: [{ type: "text", text: "test" }] });
+
+    const usageUpdate = updates.find((u: any) => u.update?.sessionUpdate === "usage_update");
+    expect(usageUpdate).toBeDefined();
+    // size should be 1000000 (Opus), not 200000 (min of both)
+    expect(usageUpdate.update.size).toBe(1000000);
+  });
+
+  it("after model switch, size updates to the new model's window", async () => {
+    const { agent, updates } = createMockAgentWithCapture();
+    // Simulate: assistant on Sonnet with both models in modelUsage
+    injectSession(agent, [
+      createAssistantMessage({ model: "claude-sonnet-4-20250514" }),
+      createResultMessageWithModel({
+        modelUsage: {
+          "claude-opus-4-20250514": {
+            inputTokens: 100,
+            outputTokens: 50,
+            cacheReadInputTokens: 20,
+            cacheCreationInputTokens: 10,
+            webSearchRequests: 0,
+            costUSD: 0.01,
+            contextWindow: 1000000,
+            maxOutputTokens: 16384,
+          },
+          "claude-sonnet-4-20250514": {
+            inputTokens: 50,
+            outputTokens: 25,
+            cacheReadInputTokens: 10,
+            cacheCreationInputTokens: 5,
+            webSearchRequests: 0,
+            costUSD: 0.005,
+            contextWindow: 200000,
+            maxOutputTokens: 16384,
+          },
+        },
+      }),
+      { type: "system", subtype: "session_state_changed", state: "idle" },
+    ]);
+
+    await agent.prompt({ sessionId: "test-session", prompt: [{ type: "text", text: "test" }] });
+
+    const usageUpdate = updates.find((u: any) => u.update?.sessionUpdate === "usage_update");
+    expect(usageUpdate).toBeDefined();
+    // size should be 200000 (Sonnet - the current model)
+    expect(usageUpdate.update.size).toBe(200000);
+  });
+
+  it("after switching back to original model, size returns to original window", async () => {
+    const { agent, updates } = createMockAgentWithCapture();
+    // Last assistant message is Opus again
+    injectSession(agent, [
+      createAssistantMessage({ model: "claude-sonnet-4-20250514" }),
+      createAssistantMessage({ model: "claude-opus-4-20250514" }),
+      createResultMessageWithModel({
+        modelUsage: {
+          "claude-opus-4-20250514": {
+            inputTokens: 200,
+            outputTokens: 100,
+            cacheReadInputTokens: 40,
+            cacheCreationInputTokens: 20,
+            webSearchRequests: 0,
+            costUSD: 0.02,
+            contextWindow: 1000000,
+            maxOutputTokens: 16384,
+          },
+          "claude-sonnet-4-20250514": {
+            inputTokens: 50,
+            outputTokens: 25,
+            cacheReadInputTokens: 10,
+            cacheCreationInputTokens: 5,
+            webSearchRequests: 0,
+            costUSD: 0.005,
+            contextWindow: 200000,
+            maxOutputTokens: 16384,
+          },
+        },
+      }),
+      { type: "system", subtype: "session_state_changed", state: "idle" },
+    ]);
+
+    await agent.prompt({ sessionId: "test-session", prompt: [{ type: "text", text: "test" }] });
+
+    const usageUpdate = updates.find((u: any) => u.update?.sessionUpdate === "usage_update");
+    expect(usageUpdate).toBeDefined();
+    // size should be 1000000 (Opus - switched back)
+    expect(usageUpdate.update.size).toBe(1000000);
+  });
+
+  it("subagent assistant messages do not affect size (top-level model is used)", async () => {
+    const { agent, updates } = createMockAgentWithCapture();
+    // Top-level assistant on Opus, then subagent on Haiku (parent_tool_use_id set)
+    injectSession(agent, [
+      createAssistantMessage({ model: "claude-opus-4-20250514" }),
+      {
+        type: "assistant" as const,
+        parent_tool_use_id: "tool_use_123",
+        uuid: randomUUID(),
+        session_id: "test-session",
+        message: {
+          model: "claude-haiku-4-5-20251001",
+          content: [{ type: "text", text: "subagent response" }],
+          usage: {
+            input_tokens: 50,
+            output_tokens: 25,
+            cache_read_input_tokens: 0,
+            cache_creation_input_tokens: 0,
+          },
+        },
+      },
+      createResultMessageWithModel({
+        modelUsage: {
+          "claude-opus-4-20250514": {
+            inputTokens: 100,
+            outputTokens: 50,
+            cacheReadInputTokens: 20,
+            cacheCreationInputTokens: 10,
+            webSearchRequests: 0,
+            costUSD: 0.01,
+            contextWindow: 1000000,
+            maxOutputTokens: 16384,
+          },
+          "claude-haiku-4-5-20251001": {
+            inputTokens: 50,
+            outputTokens: 25,
+            cacheReadInputTokens: 0,
+            cacheCreationInputTokens: 0,
+            webSearchRequests: 0,
+            costUSD: 0.001,
+            contextWindow: 200000,
+            maxOutputTokens: 8192,
+          },
+        },
+      }),
+      { type: "system", subtype: "session_state_changed", state: "idle" },
+    ]);
+
+    await agent.prompt({ sessionId: "test-session", prompt: [{ type: "text", text: "test" }] });
+
+    const usageUpdate = updates.find((u: any) => u.update?.sessionUpdate === "usage_update");
+    expect(usageUpdate).toBeDefined();
+    // size should be 1000000 (Opus - the top-level model), NOT 200000 (Haiku subagent)
+    expect(usageUpdate.update.size).toBe(1000000);
+  });
+
+  it("prefix-matches when assistant model has date suffix but modelUsage key does not", async () => {
+    const { agent, updates } = createMockAgentWithCapture();
+    // The API response has the full versioned model ID on assistant messages,
+    // but the SDK's streaming path may key modelUsage by the shorter alias.
+    injectSession(agent, [
+      createAssistantMessage({ model: "claude-opus-4-6-20250514" }),
+      createResultMessageWithModel({
+        modelUsage: {
+          "claude-opus-4-6": {
+            inputTokens: 100,
+            outputTokens: 50,
+            cacheReadInputTokens: 20,
+            cacheCreationInputTokens: 10,
+            webSearchRequests: 0,
+            costUSD: 0.01,
+            contextWindow: 1000000,
+            maxOutputTokens: 16384,
+          },
+        },
+      }),
+      { type: "system", subtype: "session_state_changed", state: "idle" },
+    ]);
+
+    await agent.prompt({ sessionId: "test-session", prompt: [{ type: "text", text: "test" }] });
+
+    const usageUpdate = updates.find((u: any) => u.update?.sessionUpdate === "usage_update");
+    expect(usageUpdate).toBeDefined();
+    // Should match via prefix: "claude-opus-4-6-20250514".startsWith("claude-opus-4-6")
+    expect(usageUpdate.update.size).toBe(1000000);
+  });
+
+  it("prefix-matches when modelUsage key has date suffix but assistant model does not", async () => {
+    const { agent, updates } = createMockAgentWithCapture();
+    injectSession(agent, [
+      createAssistantMessage({ model: "claude-opus-4-6" }),
+      createResultMessageWithModel({
+        modelUsage: {
+          "claude-opus-4-6-20250514": {
+            inputTokens: 100,
+            outputTokens: 50,
+            cacheReadInputTokens: 20,
+            cacheCreationInputTokens: 10,
+            webSearchRequests: 0,
+            costUSD: 0.01,
+            contextWindow: 1000000,
+            maxOutputTokens: 16384,
+          },
+        },
+      }),
+      { type: "system", subtype: "session_state_changed", state: "idle" },
+    ]);
+
+    await agent.prompt({ sessionId: "test-session", prompt: [{ type: "text", text: "test" }] });
+
+    const usageUpdate = updates.find((u: any) => u.update?.sessionUpdate === "usage_update");
+    expect(usageUpdate).toBeDefined();
+    expect(usageUpdate.update.size).toBe(1000000);
+  });
+
+  it("synthetic assistant messages do not override lastAssistantModel", async () => {
+    const { agent, updates } = createMockAgentWithCapture();
+    // Real assistant on Opus, then a synthetic message (e.g. from /compact)
+    injectSession(agent, [
+      createAssistantMessage({ model: "claude-opus-4-20250514" }),
+      {
+        type: "assistant" as const,
+        parent_tool_use_id: null,
+        uuid: randomUUID(),
+        session_id: "test-session",
+        message: {
+          model: "<synthetic>",
+          content: [{ type: "text", text: "compacted" }],
+          usage: {
+            input_tokens: 0,
+            output_tokens: 0,
+            cache_read_input_tokens: 0,
+            cache_creation_input_tokens: 0,
+          },
+        },
+      },
+      createResultMessageWithModel({
+        modelUsage: {
+          "claude-opus-4-20250514": {
+            inputTokens: 100,
+            outputTokens: 50,
+            cacheReadInputTokens: 20,
+            cacheCreationInputTokens: 10,
+            webSearchRequests: 0,
+            costUSD: 0.01,
+            contextWindow: 1000000,
+            maxOutputTokens: 16384,
+          },
+        },
+      }),
+      { type: "system", subtype: "session_state_changed", state: "idle" },
+    ]);
+
+    await agent.prompt({ sessionId: "test-session", prompt: [{ type: "text", text: "test" }] });
+
+    const usageUpdate = updates.find((u: any) => u.update?.sessionUpdate === "usage_update");
+    expect(usageUpdate).toBeDefined();
+    // size should be 1000000 (Opus), not 200000 (the fallback if <synthetic> overrode the model)
+    expect(usageUpdate.update.size).toBe(1000000);
+  });
+});

src/tests/authorization.test.ts

@@ -123,14 +123,17 @@ describe("authorization", () => {
       } as any,
     });
     expect(initializeResponse.authMethods).not.toContainEqual(
-      expect.objectContaining({ id: "claude-login" }),
+      expect.objectContaining({ id: "claude-ai-login" }),
+    );
+    expect(initializeResponse.authMethods).not.toContainEqual(
+      expect.objectContaining({ id: "console-login" }),
     );
     expect(initializeResponse.authMethods).toContainEqual(
       expect.objectContaining({ id: "gateway" }),
     );
   });
 
-  it("hide terminal auth even when terminal-auth is set", async () => {
+  it("hide claude auth but still show console login when terminal-auth is set", async () => {
     const [agent] = await createAgentMock();
     vi.stubGlobal("process", { ...process, argv: ["--hide-claude-auth"] });
 
@@ -140,7 +143,30 @@ describe("authorization", () => {
         _meta: { "terminal-auth": true },
       },
     });
-    expect(initializeResponse.authMethods).toStrictEqual([]);
+    expect(initializeResponse.authMethods).not.toContainEqual(
+      expect.objectContaining({ id: "claude-ai-login" }),
+    );
+    expect(initializeResponse.authMethods).toContainEqual(
+      expect.objectContaining({ id: "console-login" }),
+    );
+  });
+
+  it("hide claude auth but still show console login with terminal capability", async () => {
+    const [agent] = await createAgentMock();
+    vi.stubGlobal("process", { ...process, argv: ["--hide-claude-auth"] });
+
+    const initializeResponse = await agent.initialize({
+      protocolVersion: 1,
+      clientCapabilities: {
+        auth: { terminal: true },
+      },
+    });
+    expect(initializeResponse.authMethods).not.toContainEqual(
+      expect.objectContaining({ id: "claude-ai-login" }),
+    );
+    expect(initializeResponse.authMethods).toContainEqual(
+      expect.objectContaining({ id: "console-login" }),
+    );
   });
 
   it("show claude authentication", async () => {
@@ -152,7 +178,10 @@ describe("authorization", () => {
     });
 
     expect(initializeResponse.authMethods).toContainEqual(
-      expect.objectContaining({ id: "claude-login" }),
+      expect.objectContaining({ id: "claude-ai-login" }),
+    );
+    expect(initializeResponse.authMethods).toContainEqual(
+      expect.objectContaining({ id: "console-login" }),
     );
   });
 });