IT之家 11 月 7 日消息,AI 编程协作工具 Amp 昨日(11 月 6 日)宣布推出 Spaces,希望构建面向 AI 智能体(Agent)和团队协作的全新工作方式。
对于使用者而言,不再需要手动创建、命名和管理多个 Agent 线程,只需对话一个主 Agent,由它理解需要、拆解任务,并让其自动生成和协调其他 Agent 线程。IT之家附上相关截图如下:
Amp 指出在传统多 Agent 工作流中,使用者往往要自己决定拆分、分派以及追踪任务,Spaces 将这一层组织工作交给 Agent:每个 Space 是使用者与单个 Agent 的对话,该 Agent 会为使用者启动并协调多个并行线程。
例如,团队可以在一个 Space 中提出“推进某项能力上线”或“开展一次安全审查”。主 Agent 会将目标转化为多个具体任务,分别创建线程执行,并将进展以实时卡片形式展现在聊天中。
Spaces 不只是单人自动化工具,它被外观为团队共同规划和指挥工作的场所:成员可在同一聊天中提出需要、反馈难题或补充想法;每项被拆出的工作会表明当前状态,以及是否需要人工介入。
使用者点击任务卡片即可进入对应线程,查看 Agent 的产出、继续迭代或直接推进交付。团队成员还能在消息下评论讨论;只有当有人借助 @amp 提及时,Amp 才会读取相关评论,以减少不必要的干扰。
Amp 将 Spaces 定位为一种“更高层级”的多 Agent 协作方式,使用者从逐条安排 Agent,转向一个具备协调能力的 Agent 描述目标,再由后者负责任务分解、线程创建、进度组织和结果汇总。
官方示例表明,一名成员曾借助单个 Space 完成服务账号能力的发布;另一个情形则是定期安全审查。使用者只需在侧边栏的 Spaces 区域点击“+”,设置标题和项目,并向 Agent 说明要处理的工作即可开始。
这一能力反映出 AI 编程工具正从“单人问答”走向“团队化、任务编排化”。当 Agent 可以自主拆解和并行执行任务时,开发者的角色可能更多转向定义目标、审查结果、处理关键决策,而非手动调度每一个子任务。
Amp 是一款面向软件开发的 AI 编程 Agent 与开发环境,核心目标是让开发者把“做什么”交给 Agent,而由 Agent 负责读代码、改代码、跑测试、迭代,并在云端或本机持续执行任务。
Amp 可以在使用者的工作区中运行工具和 Shell 命令,用于检查代码、执行测试并快速迭代;它还能借助 Task 工具把特定子任务交给子 Agent 处理。
本文来源:https://www.ithome.com/1/023/609.htm
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
Thanks to IT之家 reader 惊讶仙人 for the tip!
IT之家, November 7 — AI coding collaboration tool Amp announced yesterday (November 6) that it has launched Spaces, aiming to build a new way of working for AI agents and team collaboration.
For users, there is no longer a need to manually create, name, and manage multiple Agent threads—just talk to one main Agent, which will understand the needs, break down tasks, and automatically generate and coordinate other Agent threads. IT之家 has attached relevant screenshots below:
Amp noted that in traditional multi-Agent workflows, users often have to decide how to split, assign, and track tasks themselves; Spaces hands this layer of organizational work to the Agent: each Space is a conversation between the user and a single Agent, and that Agent launches and coordinates multiple parallel threads for the user.
For example, a team can say in one Space, “drive a capability to launch” or “run a security review.” The main Agent will turn the goal into multiple concrete tasks, create separate threads to execute them, and surface progress as real-time cards in the chat.
Spaces is not just a single-user automation tool—it is presented as a place for teams to jointly plan and direct work: members can raise needs, flag difficulties, or add ideas in the same chat; each piece of work broken out will show its current status and whether human intervention is required.
Users can click a task card to enter the corresponding thread, view the Agent’s output, continue iterating, or directly push it toward delivery. Team members can also comment and discuss under messages; Amp will read relevant comments only when someone mentions it via @amp, to reduce unnecessary noise.
Amp positions Spaces as a “higher-level” form of multi-Agent collaboration, shifting users from assigning Agents command by command to describing goals to an Agent that can coordinate, with that Agent then responsible for task decomposition, thread creation, progress organization, and result aggregation.
Official examples show that one member used a single Space to complete the release of a service-account capability; another scenario is a periodic security review. Users can start simply by clicking “+” in the Spaces area of the sidebar, setting a title and project, and telling the Agent what work to handle.
This capability reflects that AI coding tools are moving from “single-user Q&A” toward “team-based, task-orchestrated” workflows. When Agents can autonomously break down and execute tasks in parallel, developers’ roles may increasingly shift to defining goals, reviewing results, and making key decisions, rather than manually scheduling every subtask.
Amp is an AI coding Agent and development environment for software development; its core goal is to let developers hand “what to do” to the Agent, while the Agent takes charge of reading code, modifying code, running tests, iterating, and continuously executing tasks in the cloud or locally.
Amp can run tools and Shell commands in the user’s workspace to check code, run tests, and iterate quickly; it can also use the Task tool to hand specific subtasks to sub-Agents for processing.


