Integration Docs - TokenClub

Claude Code is Anthropic's official CLI coding assistant, bringing Claude models into your terminal to read and write code files, run commands, and perform Git operations directly. It's one of today's most popular terminal AI coding tools. We recommend adding the settings to the env field of ~/.claude/settings.json so they apply globally across all projects. On Windows the path is C:\Users\<username>\.claude\settings.json.Claude Code 是 Anthropic 官方 CLI 編程助手，將 Claude 模型整合到終端中，可直接讀寫程式碼檔案、執行命令與 Git 操作，是目前最主流的終端 AI 編程工具之一。推薦寫入 ~/.claude/settings.json 的 env 欄位，所有專案全域生效。Windows 路徑為 C:\Users\<使用者名稱>\.claude\settings.json。

~/.claude/settings.json

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.tokenclub.cc",
    "ANTHROPIC_AUTH_TOKEN": "your-token你的令牌"
  }
}

Or use environment variables (writing them into your shell config makes them permanent):或使用環境變數方式（寫入 shell 設定可永久生效）：

bash / zsh

export ANTHROPIC_BASE_URL=https://api.tokenclub.cc
export ANTHROPIC_AUTH_TOKEN=your-token你的令牌

Note注意:：
① Do not add a trailing slash to the Base URL, otherwise path concatenation produces a double slash and triggers an error.
② When using a third-party API, set hasCompletedOnboarding in ~/.claude.json to true to skip Anthropic's official login flow.
③ After changing the config, open a new terminal window before running claude, and use /status to confirm the endpoint points to TokenClub.① Base URL 末尾不要加斜線，否則拼接路徑會出現雙斜線導致報錯。
② 使用第三方 API 時，需將 ~/.claude.json 的 hasCompletedOnboarding 設為 true，以跳過 Anthropic 官方登入引導。
③ 修改設定後請新開終端視窗再執行 claude，並可用 /status 確認位址已指向 TokenClub。

⚡ One-click import to CC Switch⚡ 一鍵導入 CC Switch

Paste your API token to generate a personalized import link; clicking it launches your local CC Switch app and adds TokenClub as a provider automatically.粘貼你的 API 令牌，生成專屬導入連結，點擊即可喚起本地 CC Switch 應用自動添加 TokenClub 供應商。

Your token is assembled into a link locally in your browser and is never uploaded to any server.
Don't have CC Switch? Get it at ccswitch.io (Windows / macOS / Linux).令牌僅在你的瀏覽器本地拼接為連結，不會上傳到任何伺服器。
未安裝 CC Switch？前往 ccswitch.io 下載（支援 Windows / macOS / Linux）。

Claude Code FAQClaude Code 常見問題

The official login flow wasn't skipped. Edit ~/.claude.json and set hasCompletedOnboarding to true, then confirm the Base URL in /status points to TokenClub. Open a new terminal after editing.未跳過官方登入引導。編輯 ~/.claude.json，將 hasCompletedOnboarding 設為 true；並確認 /status 中 Base URL 已指向 TokenClub。修改後需新開終端。

There's an extra trailing slash on the Base URL. Change it to https://api.tokenclub.cc (no trailing /).Base URL 末尾多了斜線。請改為 https://api.tokenclub.cc（結尾無 /）。

Some models enable thinking mode by default and require a thinking field in the response. Disable thinking mode in the group / model settings, or switch to a model that doesn't force thinking.部分模型預設開啟思考模式，要求回傳 thinking 欄位。可在分組 / 模型設定中關閉思考模式，或改用未強制思考的模型。

Codex CLI is OpenAI's open-source command-line AI coding assistant that lets you generate code, fix bugs, and scaffold projects in your terminal using natural language. Edit ~/.codex/config.toml, add a provider entry, and set it as the default. The token is supplied via the environment variable named in env_key.Codex CLI 是 OpenAI 開源的命令列 AI 編程助手，可在終端中透過自然語言產生程式碼、修復 bug、建構專案。編輯 ~/.codex/config.toml，新增一個 provider 節點並設為預設。令牌透過 env_key 指定的環境變數提供。

~/.codex/config.toml

model = "your-model-name你要使用的模型名"
model_provider = "tokenclub"

[model_providers.tokenclub]
name = "TokenClub"
base_url = "https://api.tokenclub.cc/v1"
wire_api = "responses"
requires_openai_auth = false
env_key = "TOKENCLUB_API_KEY"

Note注意:：
① Set the environment variable before launching: export TOKENCLUB_API_KEY=your-token.
② Codex uses the Responses API by default. If a call fails with an unsupported-protocol error, change wire_api to "chat".
③ Keys like model_provider / base_url must live in the user-level config (~/.codex/config.toml); placing them in a project-level .codex/config.toml is ignored.① 啟動前設定環境變數：export TOKENCLUB_API_KEY=你的令牌。
② Codex 預設使用 Responses API。若調用報錯提示協議不支援，將 wire_api 改為 "chat"。
③ model_provider / base_url 等鍵必須放在使用者級設定（~/.codex/config.toml），放在專案級 .codex/config.toml 會被忽略。

⚡ One-click import to CC Switch⚡ 一鍵導入 CC Switch

Paste your API token to generate a personalized import link; clicking it launches your local CC Switch app and adds TokenClub as a provider automatically.粘貼你的 API 令牌，生成專屬導入連結，點擊即可喚起本地 CC Switch 應用自動添加 TokenClub 供應商。

Your token is assembled into a link locally in your browser and is never uploaded to any server.
Don't have CC Switch? Get it at ccswitch.io (Windows / macOS / Linux).令牌僅在你的瀏覽器本地拼接為連結，不會上傳到任何伺服器。
未安裝 CC Switch？前往 ccswitch.io 下載（支援 Windows / macOS / Linux）。

Codex CLI FAQCodex CLI 常見問題

The model or group doesn't have Responses protocol conversion enabled. Change wire_api from "responses" to "chat".該模型或分組未啟用 Responses 協議轉換。將 wire_api 由 "responses" 改為 "chat" 即可。

The provider and base_url must be in the user-level ~/.codex/config.toml; these keys in a project-level .codex/config.toml are ignored.provider 與 base_url 必須寫在使用者級 ~/.codex/config.toml，專案級 .codex/config.toml 中的這些鍵會被忽略。

Codex relies on the model's function / tool-calling ability, which some smaller models lack. Choose a model that supports tool calling (e.g. GPT-5.x, Claude, Qwen Coder 32B+).Codex 依賴模型的函式 / 工具調用能力，部分較小模型不支援。請選用支援工具調用的模型（如 GPT-5.x、Claude、Qwen Coder 32B+）。

OpenCode is an open-source terminal AI coding agent (30k+ GitHub stars) supporting Claude, GPT-4, Gemini and many other models, letting you write code and fix bugs in your terminal with natural language. It uses an Anthropic-compatible protocol; we recommend writing to ~/.config/opencode/settings.json or configuring interactively with the /connect command.OpenCode 是一個開源的終端 AI 編程代理（GitHub 3 萬+ 星標），支援 Claude、GPT-4、Gemini 等多種模型，讓你在終端中用自然語言寫程式碼、修 Bug。使用 Anthropic 相容協議，推薦寫入 ~/.config/opencode/settings.json 或使用 /connect 命令互動式設定。

~/.config/opencode/settings.json

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.tokenclub.cc",
    "ANTHROPIC_AUTH_TOKEN": "your-token你的令牌"
  }
}

Or configure via the interactive command (v1.14.24+):或透過互動命令設定（v1.14.24+）：

OpenCode

opencode
# In OpenCode, type:# 在 OpenCode 介面輸入：
/connect
# Enter your API token and choose the model# 輸入你的 API 令牌，選擇對應模型

Note注意:：
① Do not add a trailing slash to the Base URL.
② When using a third-party API, set hasCompletedOnboarding to true to skip Anthropic's official login flow.
③ Open a new terminal after changing the config before running it.① Base URL 末尾不要加斜線。
② 使用第三方 API 時需設 hasCompletedOnboarding 為 true，跳過 Anthropic 官方登入引導。
③ 修改設定後需新開終端再執行。

⚡ One-click import to CC Switch⚡ 一鍵導入 CC Switch

Paste your API token to generate a personalized import link; clicking it launches your local CC Switch app and adds TokenClub as a provider automatically.粘貼你的 API 令牌，生成專屬導入連結，點擊即可喚起本地 CC Switch 應用自動添加 TokenClub 供應商。

Your token is assembled into a link locally in your browser and is never uploaded to any server.
Don't have CC Switch? Get it at ccswitch.io (Windows / macOS / Linux).令牌僅在你的瀏覽器本地拼接為連結，不會上傳到任何伺服器。
未安裝 CC Switch？前往 ccswitch.io 下載（支援 Windows / macOS / Linux）。

OpenCode FAQOpenCode 常見問題

Edit ~/.opencode.json (or a similar path), set hasCompletedOnboarding to true, and confirm the Base URL points to TokenClub. Open a new terminal afterward.編輯 ~/.opencode.json（或類似路徑），將 hasCompletedOnboarding 設為 true，並確認 Base URL 已指向 TokenClub。設定後新開終端。

OpenCode natively uses the Anthropic protocol. To use an OpenAI-compatible endpoint, convert via a proxy such as LiteLLM, or search the OpenCode community for third-party proxy solutions.OpenCode 原生使用 Anthropic 協議。若需使用 OpenAI 相容介面，可透過 LiteLLM 等代理轉換，或在 OpenCode 社群中搜尋第三方 proxy 方案。

Upgrade OpenCode to v1.14.24 or later. Older versions require manually editing settings.json.請升級 OpenCode 至 v1.14.24 或更高版本。舊版本需手動編輯 settings.json 設定。

OpenClaw is an open-source personal AI agent gateway supporting both the Anthropic Messages API and the OpenAI Completions API, letting you manage multiple model providers and orchestrate tool chains locally. We recommend configuring a custom provider in ~/.openclaw/config.json; connecting via the native Anthropic format unlocks full features like Prompt Caching and Extended Thinking.OpenClaw 是一個開源的個人 AI 智慧體閘道，支援 Anthropic Messages API 與 OpenAI Completions API 雙協議，可在本地統一管理多個模型供應商、編排工具鏈。推薦使用 ~/.openclaw/config.json 設定自訂 provider，以 Anthropic 原生格式接入可獲得 Prompt Caching 與 Extended Thinking 等完整功能。

~/.openclaw/config.json

{
  "models": {
    "mode": "merge",
    "providers": {
      "tokenclub": {
        "baseUrl": "https://api.tokenclub.cc",
        "apiKey": "your-token你的令牌",
        "api": "anthropic-messages",
        "headers": {
          "anthropic-version": "2023-06-01"
        },
        "models": [
          {
            "id": "claude-sonnet-4-6",
            "name": "Claude Sonnet 4.6",
            "contextWindow": 200000,
            "maxTokens": 16384
          }
        ]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": { "primary": "tokenclub/claude-sonnet-4-6" }
    }
  }
}

To use the OpenAI-compatible format, change api to "openai-completions" and append /v1 to the Base URL (note: Prompt Caching and Extended Thinking are unavailable in this mode):如使用 OpenAI 相容格式，將 api 改為 "openai-completions"，Base URL 末尾加 /v1（注意：該模式下 Prompt Caching 與 Extended Thinking 不可用）：

OpenAI-compatible snippetOpenAI 相容模式片段

"tokenclub": {
  "baseUrl": "https://api.tokenclub.cc/v1",
  "apiKey": "your-token你的令牌",
  "api": "openai-completions",
  "models": [
    { "id": "deepseek-chat", "name": "DeepSeek Chat" }
  ]
}

Note注意:：
① For the native Anthropic format, do not add /v1 to the Base URL, and include the anthropic-version request header.
② api can be "anthropic-messages" (recommended, supports Prompt Caching) or "openai-completions" (no /v1 restriction).
③ After editing the config, run openclaw gateway restart to apply it, and verify with openclaw models list.
④ The model id must match the name in the console model marketplace.① Anthropic 原生格式 Base URL不要加 /v1，並需攜帶 anthropic-version 請求標頭。
② api 可選 "anthropic-messages"（推薦，支援 Prompt Caching）或 "openai-completions"（無 /v1 限制）。
③ 設定修改後執行 openclaw gateway restart 使其生效，可用 openclaw models list 驗證。
④ 模型 id 需與控制台模型廣場中的名稱一致。

⚡ One-click import to CC Switch⚡ 一鍵導入 CC Switch

Paste your API token to generate a personalized import link; clicking it launches your local CC Switch app and adds TokenClub as a provider automatically.粘貼你的 API 令牌，生成專屬導入連結，點擊即可喚起本地 CC Switch 應用自動添加 TokenClub 供應商。

Your token is assembled into a link locally in your browser and is never uploaded to any server.
Don't have CC Switch? Get it at ccswitch.io (Windows / macOS / Linux).令牌僅在你的瀏覽器本地拼接為連結，不會上傳到任何伺服器。
未安裝 CC Switch？前往 ccswitch.io 下載（支援 Windows / macOS / Linux）。

OpenClaw FAQOpenClaw 常見問題

We recommend "api": "anthropic-messages" for Prompt Caching (lowering long-conversation cost) and Extended Thinking. Use "openai-completions" only when calling non-Claude models.推薦 "api": "anthropic-messages"，可獲得 Prompt Caching（降低長對話成本）和 Extended Thinking 能力。僅調用非 Claude 模型時用 "openai-completions"。

Run openclaw gateway restart to restart the gateway, then openclaw models list to check. Confirm the provider name matches the prefix in agents.defaults.model.primary.執行 openclaw gateway restart 重啟閘道，然後 openclaw models list 查看。確認 provider 名稱與 agents.defaults.model.primary 中的前綴一致。

With "api": "anthropic-messages", do not add /v1 (use https://api.tokenclub.cc). With "api": "openai-completions", you need /v1 (use https://api.tokenclub.cc/v1)."api": "anthropic-messages" 時不加 /v1（填 https://api.tokenclub.cc）。"api": "openai-completions" 時需要 /v1（填 https://api.tokenclub.cc/v1）。

WorkBuddy is Tencent's AI desktop agent workspace where you assign tasks in natural language and let it autonomously handle file processing, document generation, and other office work. It uses ~/.workbuddy/models.json to configure custom models, and the URL must end with /v1/chat/completions.WorkBuddy 是騰訊推出的 AI 桌面智慧體工作台，使用者可用自然語言下達任務，讓它自主完成檔案處理、文檔產生等辦公操作。它使用 ~/.workbuddy/models.json 設定自訂模型，URL 必須以 /v1/chat/completions 結尾。

~/.workbuddy/models.json

{
  "models": [
    {
      "id": "your-model-name你的模型名",
      "name": "TokenClub",
      "vendor": "TokenClub",
      "apiKey": "your-token你的令牌",
      "url": "https://api.tokenclub.cc/v1/chat/completions",
      "supportsToolCall": true,
      "maxInputTokens": 128000,
      "maxOutputTokens": 8192
    }
  ],
  "availableModels": ["your-model-name你的模型名"]
}

Note注意:：
① After configuring, you must fully quit and restart WorkBuddy (on Windows, exit from the system tray; on macOS, right-click the Dock icon and quit), otherwise the config won't take effect.
② id must exactly match the model name in the console model marketplace (case-sensitive).
③ For Agent mode (function calling), supportsToolCall must be true.① 設定完成後需完全退出並重啟 WorkBuddy（Windows 從系統匣退出，macOS 右鍵 Dock 圖示退出），否則設定不生效。
② id 必須與控制台模型廣場中的模型名完全一致（區分大小寫）。
③ 如需 Agent 模式（函式調用），supportsToolCall 必須為 true。

WorkBuddy FAQWorkBuddy 常見問題

You need to fully quit and restart WorkBuddy (closing the window isn't the same as quitting). Also check the JSON for errors (e.g. trailing commas, keys not wrapped in double quotes).需完全退出重啟 WorkBuddy（關閉視窗不等於退出）。同時檢查 JSON 格式是否有錯（如多餘逗號、Key 未用雙引號）。

supportsToolCall must be set to true, and the model itself must support function calling.supportsToolCall 必須設為 true，且使用的模型本身需支援函式調用能力。

The id field doesn't match the actual server-side model name. Confirm the exact model name in the console model marketplace (including hyphens and casing).id 欄位值與伺服端實際模型名不一致。請確認控制台模型廣場中的準確模型名稱（含連字號和大小寫）。

CodeBuddy is Tencent Cloud's AI coding assistant — the industry's first to support plugins (VS Code/JetBrains), a standalone IDE, and a CLI at once, deeply integrating large models like DeepSeek and Hunyuan. You can connect an OpenAI-compatible endpoint via models.json. Install: npm install -g @tencent-ai/codebuddy-code.CodeBuddy 是騰訊雲 AI 編程助手，業界首個同時支援外掛（VS Code/JetBrains）、獨立 IDE 和 CLI 三種形態，深度融合 DeepSeek/混元等大模型。透過 models.json 可接入 OpenAI 相容介面。安裝：npm install -g @tencent-ai/codebuddy-code。

~/.codebuddy/models.json

{
  "models": [
    {
      "id": "your-model-name你的模型名",
      "name": "TokenClub",
      "vendor": "TokenClub",
      "apiKey": "your-token你的令牌",
      "url": "https://api.tokenclub.cc/v1/chat/completions",
      "supportsToolCall": true
    }
  ],
  "availableModels": ["your-model-name你的模型名"]
}

Note注意:：
① Restart CodeBuddy after changing the config.
② CodeBuddy also lets you add a custom provider graphically in its settings (choose OpenAI Compatible).① 修改設定後需重啟 CodeBuddy。
② CodeBuddy 也支援在設定介面中以圖形化方式新增自訂 Provider（選擇 OpenAI Compatible）。

CodeBuddy FAQCodeBuddy 常見問題

Run npm install -g @tencent-ai/codebuddy-code, then start it with codebuddy. On first use you'll choose a login method (International / China / Enterprise).執行 npm install -g @tencent-ai/codebuddy-code，安裝後 codebuddy 啟動。首次使用需選擇登入方式（國際站 / 中國站 / 企業版）。

Not required. You can go to Settings → Model config → Add custom model, choose OpenAI Compatible as the provider, and fill in the Base URL and token.不必須。可以在設定 → 模型設定 → 新增自訂模型，Provider 選擇 OpenAI Compatible，填入 Base URL 和令牌。

It must end with /v1/chat/completions, based on the OpenAI-compatible format.必須以 /v1/chat/completions 結尾，基於 OpenAI 相容格式。

Cursor is an AI-native code editor built on VS Code with deep LLM integration, supporting intelligent code generation and multi-file editing through natural-language conversation. Connect by entering an OpenAI API Key and overriding the Base URL in Settings → Models. The Model Name must exactly match the name in the console model marketplace.Cursor 是基於 VS Code 建構的 AI 原生程式碼編輯器，深度整合大語言模型，支援透過自然語言對話實現智慧程式碼產生與多檔案編輯。在 Settings → Models 中填入 OpenAI API Key 並覆寫 Base URL 即可接入。Model Name 需嚴格匹配控制台模型廣場中的名稱。

Open Cursor → Settings → Models打開 Cursor → Settings → Models
Paste your token into the OpenAI API Key field在 OpenAI API Key 欄粘貼你的令牌
Expand Override OpenAI Base URL and enter https://api.tokenclub.cc/v1展開 Override OpenAI Base URL，填入 https://api.tokenclub.cc/v1
Enter the model name in the model list and tick it to enable在模型列表中輸入模型名並勾選啟用

Note注意:：
① Do not add /chat/completions to the Base URL — Cursor appends it automatically.
② The model name must exactly match the name in the console model marketplace.
③ Cursor's Agent Composer mode may restrict non-official-Anthropic models; Chat / Composer modes are unaffected.① Base URL不要加 /chat/completions，Cursor 會自動拼接。
② 模型名需與控制台模型廣場中的名稱完全一致。
③ Cursor Agent Composer 模式可能限制非 Anthropic 官方模型，Chat / Composer 模式不受影響。

Cursor FAQCursor 常見問題

Yes. Enter https://api.tokenclub.cc/v1 (with /v1). Don't add /chat/completions.需要。填入 https://api.tokenclub.cc/v1（帶 /v1）。不要加 /chat/completions。

Cursor's Agent Composer mode has restrictions on third-party APIs. Chat mode and regular Composer mode work normally.Cursor 的 Agent Composer 模式對第三方 API 有限制。Chat 模式和普通 Composer 模式正常可用。

The Model Name must strictly match the name in the console model marketplace (including casing and hyphens).Model Name 必須與控制台模型廣場中的名稱嚴格一致（含大小寫和連字號）。

Trae is ByteDance's AI IDE. Since v3.3.51 it natively supports a custom Base URL, so you can connect a third-party API without any extra proxy tool. Minimum version: Trae v3.3.51+. The Base URL must be the full endpoint path.Trae 是字節跳動推出的 AI IDE，v3.3.51 起原生支援自訂 Base URL，無需額外代理工具即可接入第三方 API。最低版本要求：Trae v3.3.51+。Base URL 需填寫完整介面路徑。

Open Trae → Settings → Models → Add model打開 Trae → Settings → Models → 添加模型
Set Provider to OpenAI服務商選擇 OpenAI
Set Model to Custom Model模型選擇 自訂模型 (Custom Model)
Set Model ID to the model name (e.g. deepseek-chat)Model ID填寫模型名（如 deepseek-chat）
Set API Key to your tokenAPI Key填寫你的令牌
Set Base URL to https://api.tokenclub.cc/v1Base URL填寫 https://api.tokenclub.cc/v1

Note注意:：
① The Base URL must be the full path — a domain alone won't work. For OpenAI-compatible, use https://api.tokenclub.cc/v1.
② Versions below v3.3.51 don't support native custom APIs; upgrade, or use the Cline plugin as an alternative.
③ If you choose the Anthropic provider, set the Base URL to https://api.tokenclub.cc (no /v1).① Base URL 必須填寫完整路徑，不能只填網域。OpenAI 相容填 https://api.tokenclub.cc/v1。
② 低於 v3.3.51 的版本不支援原生自訂 API，請升級或使用 Cline 外掛作為替代方案。
③ 若選擇 Anthropic 服務商，Base URL 填 https://api.tokenclub.cc（無 /v1）。

Trae FAQTrae 常見問題

Open Trae → Menu → About to see the version number. You need ≥ v3.3.51 for native custom Base URL support.打開 Trae → 選單 → 關於，查看版本號。需 ≥ v3.3.51 才支援原生自訂 Base URL。

Upgrade to the latest Trae. If you can't upgrade, install the Cline plugin (search "Cline" in Trae's plugin marketplace) and configure the custom API there.升級到最新版 Trae。如果無法升級，可安裝 Cline 外掛（Trae 外掛市場搜尋 Cline），在外掛中設定自訂 API。

Trae requires the Base URL to be the full endpoint path. For OpenAI-compatible use https://api.tokenclub.cc/v1; for Anthropic-compatible use https://api.tokenclub.cc.Trae 要求 Base URL 為完整介面路徑。OpenAI 相容填 https://api.tokenclub.cc/v1，Anthropic 相容填 https://api.tokenclub.cc。

Cline and Roo Code are VS Code plugins. Fill in the following in the plugin's settings panel:Cline 與 Roo Code 是 VS Code 外掛，在外掛設定面板中按以下步驟填寫即可：

Set API Provider to OpenAI CompatibleAPI Provider 選擇 OpenAI Compatible
Set Base URL to https://api.tokenclub.cc/v1Base URL 填 https://api.tokenclub.cc/v1
Set API Key to your tokenAPI Key 填你的令牌
Set Model ID to the specific model name you want to use (e.g. deepseek-chat)Model ID 填你要使用的具體模型名（如 deepseek-chat）

Cline / Roo Code FAQCline / Roo Code 常見問題

Choose OpenAI Compatible — don't pick Anthropic or the official OpenAI provider.選擇 OpenAI Compatible（OpenAI 相容），不要選 Anthropic 或官方 OpenAI。

The Model ID must be the exact model name from the console "Model Marketplace", case-sensitive.Model ID 需填寫控制台「模型廣場」中的準確模型名稱，區分大小寫。

The OpenAI-compatible endpoint needs /v1, i.e. https://api.tokenclub.cc/v1.OpenAI 相容介面需帶 /v1，即 https://api.tokenclub.cc/v1。

The platform is compatible with the OpenAI API protocol — just point base_url to TokenClub.平台相容 OpenAI 介面協議，只需把 base_url 指向 TokenClub 即可。

Python

from openai import OpenAI

client = OpenAI(
    base_url="https://api.tokenclub.cc/v1",
    api_key="your-token你的令牌",
)

resp = client.chat.completions.create(
    model="deepseek-chat",
    messages=[{"role": "user", "content": "Hello你好"}],
)
print(resp.choices[0].message.content)

JavaScript

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.tokenclub.cc/v1",
  apiKey: "your-token你的令牌",
});

const resp = await client.chat.completions.create({
  model: "deepseek-chat",
  messages: [{ role: "user", content: "Hello你好" }],
});
console.log(resp.choices[0].message.content);

curl

curl https://api.tokenclub.cc/v1/chat/completions \
  -H "Authorization: Bearer your-token你的令牌" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-chat",
    "messages": [{"role": "user", "content": "Hello你好"}]
  }'

SDK Integration FAQSDK 接入常見問題

The OpenAI SDK uses https://api.tokenclub.cc/v1 (with /v1). The Anthropic SDK uses https://api.tokenclub.cc (without).OpenAI SDK 使用 https://api.tokenclub.cc/v1（帶 /v1）。Anthropic SDK 使用 https://api.tokenclub.cc（不帶）。

With an SDK, just pass api_key; for raw HTTP requests, use the Authorization: Bearer your-token header.SDK 傳 api_key 即可；裸 HTTP 請求用 Authorization: Bearer 你的令牌 標頭。

Under heavy load the server sends blank lines to keep the connection alive; this doesn't affect OpenAI SDK parsing. If you parse the HTTP response yourself, ignore these blank lines.高負載時伺服器會發送空行保持連接，不影響 OpenAI SDK 解析。若自行解析 HTTP 回應，請忽略這些空行。

Status狀態碼	Meaning含義	Cause & solution原因與解決
400	Bad format格式錯誤	The request body is malformed. Fix the body format and fields per the error message.請求體格式有誤。按錯誤資訊修正請求體格式與欄位。
401	Auth failed認證失敗	Token is wrong, expired, or lacks permission. Check and update your API token.令牌錯誤、過期或無權限。檢查並更新 API 令牌。
402	Insufficient balance餘額不足	Account balance is too low. Contact TokenClub for a redemption code to top up.帳戶餘額不足。請聯絡 TokenClub 獲取兌換碼充值。
422	Parameter error參數錯誤	Invalid request body parameters. Fix the relevant parameters per the error message.請求體參數不合法。按錯誤資訊修正相關參數。
429	Rate limit reached請求達到上限	Request rate (RPM/TPM) exceeded. Lower your request frequency or retry later.請求速率（RPM/TPM）超限。降低請求頻率或稍後重試。
500	Server error伺服器故障	Internal service error. Retry later; if it persists, contact support.服務內部錯誤。稍後重試，持續出現請聯絡客服。
503	Server busy伺服器繁忙	Load is too high to process right now. Please retry later.負載過高，暫時無法處理。請稍後重試。

Quick Start快速接入