Loading...
    • 開發者指南
    • API 參考
    • MCP
    • 資源
    • 發行說明
    Search...
    ⌘K
    開始使用
    Claude 簡介快速開始
    模型與定價
    模型概覽選擇模型Claude 4.5 新功能遷移至 Claude 4.5模型棄用定價
    使用 Claude 構建
    功能概覽使用 Messages API上下文窗口提示詞最佳實踐
    功能
    提示詞快取上下文編輯擴展思考努力串流消息批次處理引用多語言支援Token 計數嵌入視覺PDF 支援Files API搜尋結果結構化輸出
    工具
    概覽如何實現工具使用細粒度工具串流Bash 工具代碼執行工具程式化工具調用計算機使用工具文字編輯器工具網頁擷取工具網頁搜尋工具記憶體工具工具搜尋工具
    Agent Skills
    概覽快速開始最佳實踐使用 API 的 Skills
    Agent SDK
    概覽快速開始TypeScript SDKTypeScript V2 (預覽)Python SDK遷移指南
    API 中的 MCP
    MCP 連接器遠端 MCP 伺服器
    第三方平台上的 Claude
    Amazon BedrockMicrosoft FoundryVertex AI
    提示詞工程
    概覽提示詞生成器使用提示詞範本提示詞改進器清晰直接使用範例 (多次提示)讓 Claude 思考 (CoT)使用 XML 標籤給 Claude 一個角色 (系統提示詞)預填 Claude 的回應鏈接複雜提示詞長上下文提示擴展思考提示
    測試與評估
    定義成功標準開發測試案例使用評估工具降低延遲
    加強防護欄
    減少幻覺增加輸出一致性緩解越獄串流拒絕減少提示詞洩露保持 Claude 的角色
    管理和監控
    Admin API 概覽使用量和成本 APIClaude Code Analytics API
    Console
    Log in
    Loading...
    Loading...
    Loading...
    Loading...
    Loading...
    Loading...
    Loading...
    Loading...
    Loading...
    Loading...
    Loading...
    Loading...
    Loading...
    Loading...
    Loading...
    Loading...

    Solutions

    • AI agents
    • Code modernization
    • Coding
    • Customer support
    • Education
    • Financial services
    • Government
    • Life sciences

    Partners

    • Amazon Bedrock
    • Google Cloud's Vertex AI

    Learn

    • Blog
    • Catalog
    • Courses
    • Use cases
    • Connectors
    • Customer stories
    • Engineering at Anthropic
    • Events
    • Powered by Claude
    • Service partners
    • Startups program

    Company

    • Anthropic
    • Careers
    • Economic Futures
    • Research
    • News
    • Responsible Scaling Policy
    • Security and compliance
    • Transparency

    Learn

    • Blog
    • Catalog
    • Courses
    • Use cases
    • Connectors
    • Customer stories
    • Engineering at Anthropic
    • Events
    • Powered by Claude
    • Service partners
    • Startups program

    Help and security

    • Availability
    • Status
    • Support
    • Discord

    Terms and policies

    • Privacy policy
    • Responsible disclosure policy
    • Terms of service: Commercial
    • Terms of service: Consumer
    • Usage policy
    工具

    使用 Claude 進行工具使用

    Claude 能夠與工具和函數互動,讓您擴展 Claude 的功能以執行更多樣的任務。
    • 使用 MCP 工具與 Claude
    • 將 MCP 工具轉換為 Claude 格式

    Claude 能夠與工具和函數互動,讓您擴展 Claude 的功能以執行更多樣的任務。

    透過我們新的課程學習掌握 Claude 工具使用所需的一切!請繼續使用此表單分享您的想法和建議。

    使用嚴格工具使用保證模式符合性

    結構化輸出為工具輸入提供有保證的模式驗證。在您的工具定義中添加 strict: true 以確保 Claude 的工具呼叫始終完全符合您的模式——不再有類型不匹配或缺少欄位的問題。

    非常適合生產代理,其中無效的工具參數會導致失敗。了解何時使用嚴格工具使用 →

    以下是如何使用 Messages API 向 Claude 提供工具的示例:

    curl https://api.anthropic.com/v1/messages \
      -H "content-type: application/json" \
      -H "x-api-key: $ANTHROPIC_API_KEY" \
      -H "anthropic-version: 2023-06-01" \
      -d '{
        "model": "claude-sonnet-4-5",
        "max_tokens": 1024,
        "tools": [
          {
            "name": "get_weather",
            "description": "Get the current weather in a given location",
            "input_schema": {
              "type": "object",
              "properties": {
                "location": {
                  "type": "string",
                  "description": "The city and state, e.g. San Francisco, CA"
                }
              },
              "required": ["location"]
            }
          }
        ],
        "messages": [
          {
            "role": "user",
            "content": "What is the weather like in San Francisco?"
          }
        ]
      }'

    工具使用的工作原理

    Claude 支援兩種類型的工具:

    1. 客戶端工具:在您的系統上執行的工具,包括:

      • 您建立和實現的使用者定義自訂工具
      • Anthropic 定義的工具,如電腦使用和文字編輯器,需要客戶端實現
    2. 伺服器工具:在 Anthropic 伺服器上執行的工具,如網路搜尋和網路擷取工具。這些工具必須在 API 請求中指定,但不需要您進行實現。

    Anthropic 定義的工具使用版本化類型(例如 web_search_20250305、text_editor_20250124)以確保跨模型版本的相容性。

    客戶端工具

    透過以下步驟將客戶端工具與 Claude 整合:

    1. 1

      向 Claude 提供工具和使用者提示

      • 在您的 API 請求中定義具有名稱、描述和輸入模式的客戶端工具。
      • 包含可能需要這些工具的使用者提示,例如「舊金山的天氣如何?」
    2. 2

      Claude 決定使用工具

      • Claude 評估任何工具是否可以幫助回答使用者的查詢。
      • 如果是,Claude 會構造一個格式正確的工具使用請求。
      • 對於客戶端工具,API 回應的 stop_reason 為 tool_use,表示 Claude 的意圖。
    3. 3

      執行工具並返回結果

      • 從 Claude 的請求中提取工具名稱和輸入
      • 在您的系統上執行工具程式碼
      • 在包含 tool_result 內容區塊的新 user 訊息中返回結果
    4. 4

      Claude 使用工具結果來制定回應

      • Claude 分析工具結果以製作對原始使用者提示的最終回應。

    注意:步驟 3 和 4 是可選的。對於某些工作流程,Claude 的工具使用請求(步驟 2)可能就是您所需的全部,無需將結果發送回 Claude。

    伺服器工具

    伺服器工具遵循不同的工作流程:

    1. 1

      向 Claude 提供工具和使用者提示

      • 伺服器工具,如網路搜尋和網路擷取,有其自己的參數。
      • 包含可能需要這些工具的使用者提示,例如「搜尋有關 AI 的最新新聞」或「分析此 URL 的內容」。
    2. 2

      Claude 執行伺服器工具

      • Claude 評估伺服器工具是否可以幫助回答使用者的查詢。
      • 如果是,Claude 執行工具,結果會自動納入 Claude 的回應中。
    3. 3

      Claude 使用伺服器工具結果來制定回應

      • Claude 分析伺服器工具結果以製作對原始使用者提示的最終回應。
      • 伺服器工具執行不需要額外的使用者互動。

    使用 MCP 工具與 Claude

    如果您正在構建使用模型上下文協議 (MCP) 的應用程式,您可以直接使用 MCP 伺服器中的工具與 Claude 的 Messages API。MCP 工具定義使用的模式格式類似於 Claude 的工具格式。您只需將 inputSchema 重新命名為 input_schema。

    不想構建自己的 MCP 客戶端? 使用 MCP 連接器直接從 Messages API 連接到遠端 MCP 伺服器,無需實現客戶端。

    將 MCP 工具轉換為 Claude 格式

    當您構建 MCP 客戶端並在 MCP 伺服器上呼叫 list_tools() 時,您將收到具有 inputSchema 欄位的工具定義。要將這些工具與 Claude 一起使用,請將它們轉換為 Claude 的格式:

    然後將這些轉換後的工具傳遞給 Claude:

    當 Claude 使用 tool_use 區塊回應時,使用 call_tool() 在您的 MCP 伺服器上執行工具,並在 tool_result 區塊中將結果返回給 Claude。

    如需構建 MCP 客戶端的完整指南,請參閱構建 MCP 客戶端。


    工具使用示例

    以下是一些代碼示例,展示了各種工具使用模式和技術。為了簡潔起見,工具是簡單的工具,工具描述比理想情況下要短,以確保最佳性能。


    定價

    Tool use requests are priced based on:

    1. The total number of input tokens sent to the model (including in the tools parameter)
    2. The number of output tokens generated
    3. For server-side tools, additional usage-based pricing (e.g., web search charges per search performed)

    Client-side tools are priced the same as any other Claude API request, while server-side tools may incur additional charges based on their specific usage.

    The additional tokens from tool use come from:

    • The tools parameter in API requests (tool names, descriptions, and schemas)
    • tool_use content blocks in API requests and responses
    • tool_result content blocks in API requests

    When you use tools, we also automatically include a special system prompt for the model which enables tool use. The number of tool use tokens required for each model are listed below (excluding the additional tokens listed above). Note that the table assumes at least 1 tool is provided. If no tools are provided, then a tool choice of none uses 0 additional system prompt tokens.

    ModelTool choiceTool use system prompt token count
    Claude Opus 4.5auto, none
    any, tool
    346 tokens
    313 tokens
    Claude Opus 4.1auto, none
    any, tool
    346 tokens
    313 tokens
    Claude Opus 4auto, none
    any, tool

    These token counts are added to your normal input and output tokens to calculate the total cost of a request.

    請參考我們的模型概覽表以了解目前的每個模型價格。

    當您發送工具使用提示時,就像任何其他 API 請求一樣,回應將輸出輸入和輸出令牌計數,作為報告的 usage 指標的一部分。


    後續步驟

    在我們的食譜中探索我們現成可實施的工具使用程式碼範例存儲庫:

    計算機工具

    了解如何將簡單的計算機工具與 Claude 整合以進行精確的數值計算。

    客戶服務代理
    from mcp import ClientSession
    
    async def get_claude_tools(mcp_session: ClientSession):
        """Convert MCP tools to Claude's tool format."""
        mcp_tools = await mcp_session.list_tools()
    
        claude_tools = []
        for tool in mcp_tools.tools:
            claude_tools.append({
                "name": tool.name,
                "description": tool.description or "",
                "input_schema": tool.inputSchema  # Rename inputSchema to input_schema
            })
    
        return claude_tools
    import anthropic
    
    client = anthropic.Anthropic()
    claude_tools = await get_claude_tools(mcp_session)
    
    response = client.messages.create(
        model="claude-sonnet-4-5",
        max_tokens=1024,
        tools=claude_tools,
        messages=[{"role": "user", "content": "What tools do you have available?"}]
    )

    346 tokens
    313 tokens
    Claude Sonnet 4.5auto, none
    any, tool
    346 tokens
    313 tokens
    Claude Sonnet 4auto, none
    any, tool
    346 tokens
    313 tokens
    Claude Sonnet 3.7 (deprecated)auto, none
    any, tool
    346 tokens
    313 tokens
    Claude Haiku 4.5auto, none
    any, tool
    346 tokens
    313 tokens
    Claude Haiku 3.5auto, none
    any, tool
    264 tokens
    340 tokens
    Claude Opus 3 (deprecated)auto, none
    any, tool
    530 tokens
    281 tokens
    Claude Sonnet 3auto, none
    any, tool
    159 tokens
    235 tokens
    Claude Haiku 3auto, none
    any, tool
    264 tokens
    340 tokens

    構建一個響應式客戶服務機器人,利用客戶端工具來增強支援。

    JSON 提取器

    查看 Claude 和工具使用如何從非結構化文本中提取結構化數據。