MoonshotAI · qipenglin · Jul 10, 2026
diff --git a/.changeset/chatgpt-oauth-provider.md b/.changeset/chatgpt-oauth-provider.md
@@ -0,0 +1,5 @@
+---
+"@moonshot-ai/kimi-code": minor
+---
+
+Add ChatGPT Plus/Pro OAuth as a login provider. Run /login and choose OpenAI ChatGPT Plus/Pro to set it up.
diff --git a/apps/kimi-code/src/tui/commands/auth.ts b/apps/kimi-code/src/tui/commands/auth.ts
@@ -1,14 +1,25 @@
+import { join } from 'node:path';
+
 import {
+  applyOpenAICodexConfig,
   applyOpenPlatformConfig,
+  FileTokenStorage,
+  fetchOpenAICodexModels,
   fetchOpenPlatformModels,
   filterModelsByPrefix,
   getOpenPlatformById,
+  loginOpenAICodexDeviceCode,
+  OPENAI_CODEX_MODELS,
+  OPENAI_CODEX_OAUTH_KEY,
+  OPENAI_CODEX_PROVIDER_NAME,
   OpenPlatformApiError,
+  resolveKimiTokenStorageName,
   type ManagedKimiCodeModelInfo,
   type ManagedKimiConfigShape,
+  type OpenAICodexModelInfo,
   type OpenPlatformDefinition,
 } from '@moonshot-ai/kimi-code-oauth';
-import { log } from '@moonshot-ai/kimi-code-sdk';
+import { log, type ModelAlias } from '@moonshot-ai/kimi-code-sdk';
 
 import type { ChoiceOption } from '../components/dialogs/choice-picker';
 import { DEFAULT_OAUTH_PROVIDER_NAME, PRODUCT_NAME } from '../constant/kimi-tui';
@@ -19,6 +30,7 @@ import {
   promptLogoutProviderSelection,
   promptModelSelectionForOpenPlatform,
   promptPlatformSelection,
+  runModelSelector,
 } from './prompts';
 import type { SlashCommandHost } from './dispatch';
 
@@ -34,12 +46,131 @@ export async function handleLoginCommand(host: SlashCommandHost): Promise<void>
     await handleKimiCodeOAuthLogin(host);
     return;
   }
+  if (platformId === OPENAI_CODEX_PROVIDER_NAME) {
+    await handleOpenAICodexOAuthLogin(host);
+    return;
+  }
 
   const platform = getOpenPlatformById(platformId);
   if (platform === undefined) return;
   await handleOpenPlatformLogin(host, platform);
 }
 
+async function handleOpenAICodexOAuthLogin(host: SlashCommandHost): Promise<void> {
+  let spinner: LoginProgressSpinnerHandle | undefined;
+  const tokenStorage = new FileTokenStorage(join(host.harness.homeDir, 'credentials'));
+  const tokenStorageName = resolveKimiTokenStorageName({
+    providerName: OPENAI_CODEX_PROVIDER_NAME,
+    oauthKey: OPENAI_CODEX_OAUTH_KEY,
+  });
+  const controller = new AbortController();
+  const cancelLogin = (): void => {
+    controller.abort();
+  };
+  host.cancelInFlight = cancelLogin;
+  let openAICodexModels: readonly OpenAICodexModelInfo[] = OPENAI_CODEX_MODELS;
+  try {
+    const token = await loginOpenAICodexDeviceCode({
+      signal: controller.signal,
+      onDeviceCode: (data) => {
+        spinner = host.showLoginAuthorizationPrompt(
+          {
+            userCode: data.userCode,
+            deviceCode: '',
+            verificationUri: data.verificationUri,
+            verificationUriComplete: data.verificationUri,
+            expiresIn: data.expiresIn,
+            interval: data.interval,
+          },
+          { title: 'Sign in to ChatGPT' },
+        );
+      },
+    });
+    await tokenStorage.save(tokenStorageName, token);
+    try {
+      openAICodexModels = await fetchOpenAICodexModels(token.accessToken, {
+        signal: controller.signal,
+      });
+    } catch (error) {
+      log.warn('failed to fetch OpenAI Codex models; using built-in fallback', {
+        providerName: OPENAI_CODEX_PROVIDER_NAME,
+        sessionId: host.session?.id,
+        error,
+      });
+    }
+    spinner?.stop({ ok: true, label: 'Logged in.' });
+    spinner = undefined;
+  } catch (error) {
+    const cancelled = controller.signal.aborted;
+    spinner?.stop({
+      ok: false,
+      label: cancelled ? 'Login cancelled.' : 'Login failed.',
+    });
+    spinner = undefined;
+    if (!cancelled) {
+      log.warn('login failed', {
+        providerName: OPENAI_CODEX_PROVIDER_NAME,
+        sessionId: host.session?.id,
+        error,
+      });
+      host.showError(`Login failed: ${formatErrorMessage(error)}`);
+    }
+    return;
+  } finally {
+    if (host.cancelInFlight === cancelLogin) {
+      host.cancelInFlight = undefined;
+    }
+  }
+
+  const modelDict: Record<string, ModelAlias> = {};
+  for (const model of openAICodexModels) {
+    const supportEfforts = [...(model.supportEfforts ?? ['low', 'medium', 'high', 'xhigh'])];
+    modelDict[`${OPENAI_CODEX_PROVIDER_NAME}/${model.id}`] = {
+      provider: OPENAI_CODEX_PROVIDER_NAME,
+      model: model.id,
+      maxContextSize: model.contextLength,
+      capabilities: model.supportsImageIn
+        ? ['tool_use', 'thinking', 'image_in']
+        : ['tool_use', 'thinking'],
+      displayName: model.displayName,
+      supportEfforts,
+      defaultEffort: model.defaultEffort ?? (supportEfforts.includes('medium') ? 'medium' : supportEfforts[0]),
+    };
+  }
+  const selection = await runModelSelector(host, modelDict);
+  if (selection === undefined) {
+    await tokenStorage.remove(tokenStorageName);
+    return;
+  }
+
+  const selectedModelId = selection.alias.slice(`${OPENAI_CODEX_PROVIDER_NAME}/`.length);
+  const existingConfig = await host.harness.getConfig();
+  if (existingConfig.providers[OPENAI_CODEX_PROVIDER_NAME] !== undefined) {
+    await host.harness.removeProvider(OPENAI_CODEX_PROVIDER_NAME);
+  }
+
+  const config = await host.harness.getConfig();
+  applyOpenAICodexConfig(config as ManagedKimiConfigShape, {
+    selectedModelId,
+    thinking: selection.thinking !== 'off',
+    effort:
+      selection.thinking !== 'off' && selection.thinking !== 'on'
+        ? selection.thinking
+        : undefined,
+    models: openAICodexModels,
+  });
+  await host.harness.setConfig({
+    providers: config.providers,
+    models: config.models,
+    defaultModel: config.defaultModel,
+    thinking: config.thinking,
+  });
+
+  await host.authFlow.refreshConfigAfterLogin();
+  host.track('login', { provider: OPENAI_CODEX_PROVIDER_NAME, method: 'oauth' });
+  host.showStatus(`Setup complete: OpenAI ChatGPT Plus/Pro · ${selectedModelId}`);
+}
+
 async function handleKimiCodeOAuthLogin(host: SlashCommandHost): Promise<void> {
   const status = await host.harness.auth.status(DEFAULT_OAUTH_PROVIDER_NAME);
   const alreadyLoggedIn = status.providers.some(
@@ -222,6 +353,14 @@ export async function handleLogoutCommand(host: SlashCommandHost): Promise<void>
   if (target === DEFAULT_OAUTH_PROVIDER_NAME) {
     await host.harness.auth.logout(DEFAULT_OAUTH_PROVIDER_NAME);
   } else {
+    if (target === OPENAI_CODEX_PROVIDER_NAME) {
+      await new FileTokenStorage(join(host.harness.homeDir, 'credentials')).remove(
+        resolveKimiTokenStorageName({
+          providerName: OPENAI_CODEX_PROVIDER_NAME,
+          oauthKey: OPENAI_CODEX_OAUTH_KEY,
+        }),
+      );
+    }
     await host.harness.removeProvider(target);
   }
 

diff --git a/apps/kimi-code/src/tui/commands/dispatch.ts b/apps/kimi-code/src/tui/commands/dispatch.ts
@@ -125,7 +125,10 @@ export interface SlashCommandHost {
 
   // UI
   showLoginProgressSpinner(label: string): LoginProgressSpinnerHandle;
-  showLoginAuthorizationPrompt(auth: DeviceAuthorization): LoginProgressSpinnerHandle;
+  showLoginAuthorizationPrompt(
+    auth: DeviceAuthorization,
+    options?: { readonly title?: string | undefined },
+  ): LoginProgressSpinnerHandle;
   showProgressSpinner(label: string): LoginProgressSpinnerHandle;
 
   // Theme

diff --git a/apps/kimi-code/src/tui/components/dialogs/platform-selector.ts b/apps/kimi-code/src/tui/components/dialogs/platform-selector.ts
@@ -1,9 +1,10 @@
-import { OPEN_PLATFORMS } from '@moonshot-ai/kimi-code-oauth';
+import { OPENAI_CODEX_PROVIDER_NAME, OPEN_PLATFORMS } from '@moonshot-ai/kimi-code-oauth';
 
 import { ChoicePickerComponent, type ChoiceOption } from './choice-picker';
 
 const PLATFORM_OPTIONS: readonly ChoiceOption[] = [
   { value: 'kimi-code', label: 'Kimi Code (OAuth)' },
+  { value: OPENAI_CODEX_PROVIDER_NAME, label: 'OpenAI ChatGPT Plus/Pro (OAuth)' },
   ...OPEN_PLATFORMS.map((platform) => ({ value: platform.id, label: platform.name })),
 ];
 

diff --git a/apps/kimi-code/src/tui/kimi-tui.ts b/apps/kimi-code/src/tui/kimi-tui.ts
@@ -2183,11 +2183,14 @@ export class KimiTUI {
     };
   }
 
-  showLoginAuthorizationPrompt(auth: DeviceAuthorization): LoginProgressSpinnerHandle {
+  showLoginAuthorizationPrompt(
+    auth: DeviceAuthorization,
+    options: { readonly title?: string | undefined } = {},
+  ): LoginProgressSpinnerHandle {
     openUrl(auth.verificationUriComplete);
     this.state.transcriptContainer.addChild(
       new DeviceCodeBoxComponent({
-        title: 'Sign in to Kimi Code',
+        title: options.title ?? 'Sign in to Kimi Code',
         url: auth.verificationUriComplete,
         code: auth.userCode,
         hint: 'Press Ctrl-C to cancel',

diff --git a/docs/en/configuration/providers.md b/docs/en/configuration/providers.md
@@ -35,7 +35,7 @@ Two paths when adding:
 - **Custom registry (api.json)**: paste a custom registry URL and Bearer token; the CLI automatically creates the `providers` / `models` entries. On later startup, providers from the same registry URL are refreshed together, so upstream provider additions, removals, and model metadata changes are synced.
 
 ::: warning
-Kimi Code OAuth managed accounts logged in via `/login` do not appear in `/provider`. Use `/login` and `/logout` to manage them.
+OAuth accounts logged in via `/login`, including Kimi Code and ChatGPT Plus/Pro, do not appear in `/provider`. Use `/login` and `/logout` to manage them.
 :::
 
 The same operations are also available in non-interactive environments via the shell command: [`kimi provider`](../reference/kimi-command.md#kimi-provider).
@@ -154,7 +154,7 @@ To route Vertex requests through a custom (e.g. proxied) endpoint, set `base_url
 
 ## OAuth and credential injection
 
-The Kimi Code managed service uses OAuth rather than static API keys. After running `/login`, the built-in authentication toolchain automatically writes and refreshes credentials — no manual configuration is needed in `config.toml` for this.
+The Kimi Code managed service and ChatGPT Plus/Pro provider use OAuth rather than static API keys. After running `/login`, the built-in authentication toolchain automatically writes and refreshes credentials — no manual configuration is needed in `config.toml` for this.
 
 ## Next steps
 

diff --git a/docs/en/guides/getting-started.md b/docs/en/guides/getting-started.md
@@ -100,9 +100,10 @@ On first launch you need to configure an API source. In the interactive UI, ente
 /login
 ```
 
-`/login` opens a platform selector supporting two options:
+`/login` opens a platform selector supporting these options:
 
 - **Kimi Code (OAuth)** — device-code flow; open the link on any device, sign in, and enter the code to authorize
+- **OpenAI ChatGPT Plus/Pro (OAuth)** — sign in with a ChatGPT subscription and choose one of the bundled Codex models
 - **Kimi Platform API key** — enter an API key from `platform.kimi.com` or `platform.kimi.ai`
 
 To sign out, enter `/logout` to clear the current credentials.

diff --git a/docs/zh/configuration/providers.md b/docs/zh/configuration/providers.md
@@ -35,7 +35,7 @@ Kimi Code CLI 支持同时接入多家 LLM 平台——用 Kimi Code 托管服
 - **Custom registry (api.json)**：粘贴自定义 registry 地址和 Bearer token，CLI 自动创建 `providers` / `models` 条目。后续启动时，同一个 registry 地址下的供应商会一起刷新，因此上游新增、删除供应商以及模型元数据变化都会同步。
 
 ::: warning
-通过 `/login` 登录的 Kimi Code OAuth 托管账号不会在 `/provider` 里显示，请用 `/login` 和 `/logout` 管理。
+通过 `/login` 登录的 OAuth 账号（包括 Kimi Code 和 ChatGPT Plus/Pro）不会在 `/provider` 里显示，请用 `/login` 和 `/logout` 管理。
 :::
 
 非交互环境下也可以用 shell 命令完成同样操作：[`kimi provider`](../reference/kimi-command.md#kimi-provider)。
@@ -154,7 +154,7 @@ kimi
 
 ## OAuth 与凭证注入
 
-Kimi Code 托管服务使用 OAuth 而非静态 API 密钥。运行 `/login` 后，内置的认证工具链会自动写入并刷新凭证，`config.toml` 里无需手动配置这部分内容。
+Kimi Code 托管服务和 ChatGPT Plus/Pro 供应商使用 OAuth 而非静态 API 密钥。运行 `/login` 后，内置的认证工具链会自动写入并刷新凭证，`config.toml` 里无需手动配置这部分内容。
 
 ## 下一步
 

diff --git a/docs/zh/guides/getting-started.md b/docs/zh/guides/getting-started.md
@@ -100,9 +100,10 @@ kimi -c
 /login
 ```
 
-`/login` 会弹出平台选择器，支持两种方式：
+`/login` 会弹出平台选择器，支持以下方式：
 
 - **Kimi Code（OAuth）** — 验证码流程，在任意设备打开链接、登录并输入验证码即可授权
+- **OpenAI ChatGPT Plus/Pro（OAuth）** — 使用 ChatGPT 订阅登录，并选择内置的 Codex 模型
 - **Kimi Platform API 密钥** — 输入来自 `platform.kimi.com` 或 `platform.kimi.ai` 的 API 密钥
 
 需要退出登录时，输入 `/logout` 清除当前凭证。

diff --git a/packages/agent-core/src/agent/compaction/full.ts b/packages/agent-core/src/agent/compaction/full.ts
@@ -407,10 +407,12 @@ export class FullCompaction {
           : undefined;
       const provider = applyCompletionBudget({
         provider: this.agent.config.provider,
-        budget: resolveCompletionBudget({
-          maxOutputSize: this.agent.config.maxOutputSize ?? defaultCompactionCap,
-          reservedContextSize: this.agent.kimiConfig?.loopControl?.reservedContextSize,
-        }),
+        budget: this.agent.config.disableCompletionBudget
+          ? undefined
+          : resolveCompletionBudget({
+              maxOutputSize: this.agent.config.maxOutputSize ?? defaultCompactionCap,
+              reservedContextSize: this.agent.kimiConfig?.loopControl?.reservedContextSize,
+            }),
         capability,
       });
       const instruction = this.buildInstruction(data.instruction);

diff --git a/packages/agent-core/src/agent/config/index.ts b/packages/agent-core/src/agent/config/index.ts
@@ -178,6 +178,10 @@ export class ConfigState {
     return this.tryResolvedProviderConfig()?.maxOutputSize;
   }
 
+  get disableCompletionBudget(): boolean {
+    return this.tryResolvedProviderConfig()?.disableCompletionBudget === true;
+  }
+
   private get resolvedProviderConfig(): ResolvedRuntimeProvider | undefined {
     if (this._modelAlias === undefined) return undefined;
     return this.agent.modelProvider?.resolveProviderConfig(this._modelAlias);

diff --git a/packages/agent-core/src/agent/index.ts b/packages/agent-core/src/agent/index.ts
@@ -307,10 +307,12 @@ export class Agent {
     // is applied in ConfigState.provider so compaction shares it. See get provider().
     const provider = this.config.provider;
     const loopControl = this.kimiConfig?.loopControl;
-    const completionBudgetConfig = resolveCompletionBudget({
-      maxOutputSize: this.config.maxOutputSize,
-      reservedContextSize: loopControl?.reservedContextSize,
-    });
+    const completionBudgetConfig = this.config.disableCompletionBudget
+      ? undefined
+      : resolveCompletionBudget({
+          maxOutputSize: this.config.maxOutputSize,
+          reservedContextSize: loopControl?.reservedContextSize,
+        });
     return new KosongLLM({
       provider,
       systemPrompt: this.config.systemPrompt,