來源Pandaily•較早收集於 55m
Floatboat 推出基於行事曆驅動工作流的「主動式代理作業系統」

#ai-agents#workflow-automation#multi-agent-systemsfloatboatfloatboatdeepseekkimilarkwechat
💡了解行事曆整合代理如何自動化行政工作流,並實現多代理團隊協作。
⚡ 30 秒速覽
有什麼變化
透過行事曆整合,自動化處理會議摘要與文件收集等重複性工作流程。
為什麼重要
此平台將 AI 從被動式聊天機器人轉向能自主管理行政負擔的主動式代理。這顯示了企業 AI 的趨勢,即將 AI 代理視為功能性團隊成員,而非僅是查詢工具。
下一步行動
評估團隊內重複性的行政任務,並測試行事曆觸發的代理程式是否能取代手動的文件收集或會議後續跟進工作。
誰應關注:Developers & AI Engineers
關鍵要點
- •透過行事曆整合,自動化處理會議摘要與文件收集等重複性工作流程。
- •FloatIM 介面支援多個 AI 代理之間的自主協作。
- •支援超過 3,500 種應用程式整合,包括 Lark 與 WeChat。
- •模型無關架構,目前運行於 DeepSeek 與 Kimi 模型。
🧠 深度解析
背景與延伸:來自公開資料,非原文內容。引用 19 個來源。
🔑 增強重點摘要
- •Floatboat, founded in 2025 by Bruce Tan and Judy Gao, is backed by venture firms Sequoia and Welight Capital.
- •The platform is built on open IACT and Selfware protocols, designed to foster an open and interoperable AI agent ecosystem.
- •Floatboat's initial offering, FloatSchedule, functions as a "work execution layer" that proactively prepares for calendar events, distinguishing it from traditional AI calendar tools focused on scheduling optimization.
- •The system is available as a free desktop application for macOS and Windows, and includes "Combo Skills" for pre-built workflow recipes.
- •Floatboat is specifically designed as an all-in-one workspace for individual entrepreneurs and small businesses, integrating a file manager, built-in browser, and AI agent into a single desktop environment.
🛠️ 技術深入
- Floatboat's architecture is built on open IACT and Selfware protocols, promoting an open and interoperable agent ecosystem.
- The FloatIM interface is a built-in messaging system that allows multiple AI agents to collaborate autonomously, treating them like team members in a group chat.
- Agents within FloatIM can hand off work, with files carrying their full context, including data, production steps, and future instructions.
- The platform supports a multi-model architecture, integrating various large language models such as DeepSeek, Kimi, GPT, Claude, and Gemini.
- Floatboat agents run locally on the user's desktop before connecting to the network, ensuring user data ownership and control.
- DeepSeek models, utilized by Floatboat, are often Mixture-of-Experts (MoE) architectures, with some variants like DeepSeek V4 Flash optimized for fast coding and agent tasks with a 1-million-token context window.
- Kimi models, also integrated, are Mixture-of-Experts (MoE) models known for large context windows (e.g., Kimi K2 with 128,000 to 256,000 tokens, and Kimi K2.5 with a 2-million-character context window in beta).
- Kimi K2.5 and K2.6 feature "Agent Swarm" technology, enabling the parallel coordination of up to 100 specialized AI agents for complex tasks.
🔮 前景展望基於引用來源的 AI 分析
Floatboat's open protocol approach will accelerate the development of an interoperable AI agent ecosystem.
By building on open IACT and Selfware protocols, Floatboat encourages third-party agent integration, fostering a broader, less siloed AI agent landscape.
The 'proactive agent OS' model will become a dominant paradigm for enterprise productivity, moving beyond reactive AI tools.
Floatboat's design to anticipate needs and automate workflows without explicit prompts represents a fundamental shift from current prompt-dependent AI assistants, potentially setting a new standard for efficiency.
Floatboat's focus on individual entrepreneurs and small businesses will democratize advanced AI agent capabilities.
By offering an all-in-one workspace for solo users, Floatboat makes sophisticated multi-agent collaboration and workflow automation accessible to a segment often underserved by enterprise-focused AI solutions.
⏳ 時間線
2023-03
Moonshot AI, developer of Kimi, founded.
2023-10
Kimi chatbot (first version) released by Moonshot AI.
2023
DeepSeek AI founded.
2025
Floatboat (proactive agent OS) founded.
2026-01
Kimi K2.5, featuring Agent Swarm technology, released by Moonshot AI.
2026-03
Floatboat completes seed round funding from Sequoia and Welight Capital.
2026-05
Floatboat globally launches its proactive agent OS and introduces FloatIM.
📎 來源 (19)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
📰
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: Pandaily ↗
每週電子報
每週一封,可隨時退訂。

