來源OpenClaw (GitHub Releases)•較早收集於 5m
OpenClaw 2026.3.28-beta.1：重大變更與新功能
#providers#plugins#image-generation#toolsopenclawopenclawxaiqwenminimaxopenai
💡xAI 工具、MiniMax 圖像、插件核准升級 OpenClaw，利於 AI 開發者（32字）
⚡ 30 秒速覽
有什麼變化
Qwen 提供者移除已棄用 OAuth，遷移至 Model Studio API 金鑰
為什麼重要
簡化提供者整合與工具工作流程，減少 xAI 及圖像生成的的手動設定。重大變更強制設定更新，但提升長期穩定性。改善跨平台 AI 應用建置，具備更佳核准與頻道支援。
下一步行動
執行 `openclaw doctor` 驗證設定並在更新至 beta.1 前遷移 Qwen 至 modelstudio-api-key。
誰應關注：Developers & AI Engineers
關鍵要點
- •Qwen 提供者移除已棄用 OAuth，遷移至 Model Studio API 金鑰
- •xAI 移至 Responses API，新增 x_search 及自動插件啟用
- •MiniMax 新增 image-01 生成與編輯，支援長寬比控制
- •插件新增非同步 requireApproval，用於工具呼叫的多頻道核准
- •OpenAI 預設啟用 apply_patch，並對齊沙盒政策
🧠 深度解析
本篇為 AI 生成分析，非原文內容。
🔑 增強重點摘要
- •The transition of the Qwen provider to Model Studio API keys aligns with Alibaba Cloud's broader strategy to unify developer access under the Model Studio (DashScope) platform, deprecating legacy OAuth flows that were previously prone to token expiration issues.
- •The integration of xAI's Responses API with x_search suggests a shift toward real-time, web-augmented reasoning capabilities within OpenClaw, moving beyond static model inference to dynamic information retrieval.
- •The introduction of async tool approval hooks for plugins addresses a critical security vulnerability in agentic workflows, allowing for human-in-the-loop verification before sensitive external API calls are executed.
🔮 前景展望基於引用來源的 AI 分析
OpenClaw will transition to a strictly API-key-based authentication model for all third-party providers.
The removal of Qwen OAuth in favor of Model Studio keys indicates a move toward standardized, centralized authentication patterns across the platform.
Plugin developers will increasingly adopt multi-channel approval workflows for agentic tool execution.
The new async requireApproval hook provides a standardized framework that reduces the risk of unauthorized tool usage in automated agent environments.
📰
AI 週報
閱讀本週精選 AI 大事摘要 →
👉相關動態
AI 策展新聞聚合。所有內容版權歸原始發布者所有。
原始來源： OpenClaw (GitHub Releases) ↗
每週電子報
每週一封，可隨時退訂。