返回新聞
產品展示AI Understanding 簡報

Anthropic 將 Claude Cowork 過渡到基於雲端的執行

Anthropic 已將 Claude Cowork 任務轉移到 Pro 和 Max 用戶的基於雲端的處理,從而實現跨裝置會話連續性,同時透過桌面應用程式保持本地文件存取。

4 min readRead the linked source
Source-page capture accompanying Anthropic transitions Claude Cowork to cloud-based execution
來源參考來源記錄
出版商
support.claude.com
來源連結
support.claude.comhttps://support.claude.com/en/articles/15520349-use-claude-cowork-on-web-desktop-and-mobile
來源類型
連結來源-主要來源狀態尚未確定。
背景60 秒內了解這一點

從這裡開始

關鍵術語

代理商工作流程
人工智慧系統規劃、執行、檢查結果並迭代實現目標的多步驟流程。
計算
訓練和運行模型所需的處理資源,通常以 FLOPS 或 GPU 小時來衡量。
特點
模型用來進行預測的輸入變數。
測試一下自己AI 代理測驗

發生了什麼事

Anthropic has transitioned Claude Cowork tasks for Pro and Max plan users to cloud-based execution, effective October 6, 2026. This shift allows users to initiate and continue Cowork sessions across web, desktop, mobile, and the Chrome side panel. While processing now occurs on Anthropic's servers, the system maintains access to local files through the Claude Desktop app, provided the application remains open on the user's machine.

As of October 6, 2026, new Claude Cowork tasks for Pro and Max subscribers are processed in the cloud rather than locally. This change eliminates the 'Only on your computer' setting in the application's general settings menu.

Existing tasks initiated on a user's local machine prior to this date will remain local until completion. Users are provided with an option to download transcripts for these legacy tasks to facilitate migration to Claude Code if desired.

The system now supports cross-surface continuity. A user can initiate a task in the Chrome side panel and resume the same session on a desktop or mobile device. Notifications for task completion or input requests are pushed to mobile devices.

Despite the cloud-based processing, the system retains the ability to interact with local folders. This requires the Claude Desktop app to be running, which acts as a bridge for the cloud-based agent to access specific local files as needed.

來源詳情: support.claude.com ↗

為什麼這很重要

This update significantly enhances the utility of Claude Cowork by decoupling task execution from a single physical machine, allowing for seamless transitions between devices. By moving processing to the cloud, Anthropic enables users to start a task on a desktop and resume it on a mobile device or browser. The integration of local file access via the desktop app balances the flexibility of cloud computing with the necessity of interacting with local development environments, though it introduces new dependencies on the desktop application's status for specific tasks. This shift marks a transition from a strictly local-first to a hybrid model, expanding the accessibility of AI-driven task automation for professional users.

The move to cloud-based execution addresses the primary limitation of local-only AI agents: the inability to maintain state across different hardware. By centralizing sessions in the cloud, Anthropic allows for a more fluid user experience that aligns with modern, multi-device professional workflows.

The hybrid approach—where the agent runs in the cloud but accesses local files via a desktop bridge—is a significant architectural choice. It allows the agent to leverage the power of Anthropic's servers while respecting the user's local file structure, provided the desktop app is active.

The transition highlights the evolving nature of AI agents from local utilities to persistent, cloud-synced assistants. This shift is critical for users who require long-running tasks that must persist even when their primary workstation is offline or inaccessible.

Interactive Mechanism

互動機制:它實際上是如何運作的

以互動方式探索這項發展背後的基礎技術。

Agent Lifecycle Stage:
1
User Intent & Planning: "Audit customer refund request #4092 and settle payment."
2
Tool Calling: Emits structured JSON call crm_get_transaction(id='4092').
3
Guardrail & Verification:🛡️ Paused: High-value action requires human operator sign-off.
4
Final Settlement: Refund recorded, email receipt dispatched, and audit log stored.
Core takeaway: An AI agent is not just a language model—it is a closed loop of planning, tool invocation, and environment feedback. Production systems require self-healing retries and strict human approval guardrails.
互動式概念檢查+10 Points
AI Agents Quiz

An agent must create a draft calendar event for Tuesday at 2 p.m. Which evidence would establish the requested result?

接下來看什麼

Users should monitor the 'beta' status of cloud-based Cowork, as Anthropic has noted that some features remain unavailable in this environment. Additionally, the removal of the 'Only on your computer' setting necessitates a shift in workflow for users who previously relied on local-only processing for privacy or security reasons. It remains to be seen how the platform will handle potential latency or connectivity issues when tasks require frequent synchronization between cloud-based agents and local file systems.

The 'beta' designation for cloud-based Cowork suggests that users may encounter limitations or gaps compared to the previous local-only implementation. Anthropic has not provided a comprehensive list of these missing features.

Data privacy remains a key area of interest. Anthropic specifies that whether conversations are used to improve AI models depends on the user's 'Help improve our AI models' privacy setting, which now applies to these cloud-based sessions.

The requirement to keep the desktop app open for local file access creates a new dependency. Users should watch for how the system handles scenarios where the desktop app is closed while a cloud-based task is attempting to access local resources.

相關指引和測驗

人工智慧代理ChatGPT 與大型語言模型AI 倫理測試你所知道的—嘗試免費的人工智慧測驗在我們的詞彙表中尋找人工智慧術語關注 AI 模型發布追蹤器
覺得有用嗎?