# The Zeroth Documentation

- [Welcome](/en/docs): Understand why The One exists for Multi-Agent architecture and where to start.
- [Quick Start](/en/docs/start/quick-start): The shortest path from installing The One to running your first Multi-Agent Graph.
- [Models and Pricing](/en/docs/start/models-and-pricing): Understand official model service, third-party providers, subscriptions, and top-ups.
- [Release Notes](/en/docs/start/release-notes): Public release notes for The One Desktop.
- [The One 0.3.0 Release Notes](/en/docs/start/release-notes/0-3-0): 0.3.0 makes agents self-sustaining — they manage their own context, grow their own capabilities, build their own Graphs, remember, and improve themselves.
- [The One 0.2.0 Release Notes](/en/docs/start/release-notes/0-2-0): 0.2.0 adds provider integration, Definition Spaces, sharing, and Graph terminology.
- [The One 0.1.0 Release Notes](/en/docs/start/release-notes/0-1-0): 0.1.0 is the first complete desktop release of The One.
- [Definition Space](/en/docs/the-one-app/definition-space): The core repository for Multi-Agent resources in The One.
- [Profile](/en/docs/the-one-app/definition-space/profile): Write, import, and reuse Prompt / Skill modules.
- [MCP](/en/docs/the-one-app/definition-space/mcp): Import and manage common MCP tools.
- [Agent](/en/docs/the-one-app/definition-space/agent): Create Agent presets with independent models, prompts, and tools.
- [Graph](/en/docs/the-one-app/definition-space/graph): Build a runnable and observable Multi-Agent Graph with nodes and edges.
- [Memory](/en/docs/the-one-app/definition-space/memory): Each Definition Space carries an agent-curated memory that travels with it.
- [Atlas](/en/docs/the-one-app/definition-space/atlas): A visual canvas for navigating and authoring the whole Definition Space.
- [Run](/en/docs/the-one-app/run): Choose a workspace path and Graph, then start a transparent Multi-Agent run.
- [Settings](/en/docs/the-one-app/settings): Manage account, Definition Spaces, API providers, and desktop runtime.
- [Space Management](/en/docs/the-one-app/settings/space-management): Create, switch, import, export, and share Definition Spaces.
- [API Configuration](/en/docs/the-one-app/settings/api-configuration): Configure official model service, third-party API keys, OAuth, and OpenAI-compatible providers.
- [Single Agent](/en/docs/multi-agent-architecture/single-agent): Understand how one Agent is determined by model, tools, and context.
- [Model](/en/docs/multi-agent-architecture/single-agent/model): The model is the Agent's brain, and different roles need different models.
- [Tools](/en/docs/multi-agent-architecture/single-agent/tools): Tools turn a language model into an Agent.
- [Context](/en/docs/multi-agent-architecture/single-agent/context): Use constraints, information, and tasks to build a clear Agent mind.
- [ReAct](/en/docs/multi-agent-architecture/single-agent/react): The One Agents run through reasoning, tool use, observation, and further reasoning.
- [Action Sequence](/en/docs/multi-agent-architecture/single-agent/action-sequence): Use action sequences to split complex goals into executable steps.
- [From Single Agent to Multi-Agent](/en/docs/multi-agent-architecture/evolution): Why complex tasks need multiple Agents with clear roles and clean context.
- [More Possibilities in Multi-Agent](/en/docs/multi-agent-architecture/more-possibilities): Capabilities that single-Agent systems cannot naturally provide.
- [Communication](/en/docs/multi-agent-architecture/more-possibilities/communication): The main ways Agents communicate in The One.
- [Context Editing](/en/docs/multi-agent-architecture/more-possibilities/context-editing): Agents can edit another Agent's context inside a Graph.
- [Dynamic Graph](/en/docs/multi-agent-architecture/more-possibilities/dynamic-graph): Agents can dynamically create, clone, activate, or suspend other Agents.
- [Meditation](/en/docs/multi-agent-architecture/more-possibilities/meditation): Agents manage their own context window by compressing their own history.
- [Self-Evolution](/en/docs/multi-agent-architecture/more-possibilities/self-evolution): An agent-only workshop that diagnoses its own failures and fixes the right layer.
- [欢迎](/zh/docs): 了解 The One 为什么为 Multi-Agent 架构而生，以及应该从哪里开始阅读。
- [快速入门](/zh/docs/start/quick-start): 从安装 The One 到运行第一个 Multi-Agent Graph 的最短路径。
- [模型与价格](/zh/docs/start/models-and-pricing): 理解官方模型服务、第三方模型渠道、订阅和用量包的边界。
- [发布记录](/zh/docs/start/release-notes): The One Desktop 的公开版本更新记录。
- [The One 0.3.0 Release Notes](/zh/docs/start/release-notes/0-3-0): 0.3.0 让 Agent 开始自我维持——自己管理上下文、自己扩展能力、自己搭建 Graph、自己记忆、自己改进。
- [The One 0.2.0 Release Notes](/zh/docs/start/release-notes/0-2-0): 0.2.0 带来 Provider 接入、Definition Space、分享平台和 Graph 语义升级。
- [The One 0.1.0 Release Notes](/zh/docs/start/release-notes/0-1-0): 0.1.0 是 The One 第一个完整桌面端公开版本。
- [仓储（Definition Space）](/zh/docs/the-one-app/definition-space): Definition Space 是 The One 管理多智能体资源的核心仓储。
- [Profile](/zh/docs/the-one-app/definition-space/profile): 编写、导入和复用 Prompt / Skill 模块。
- [MCP](/zh/docs/the-one-app/definition-space/mcp): 导入和管理常用 MCP 工具。
- [Agent](/zh/docs/the-one-app/definition-space/agent): 创建具有独立模型、提示词和工具能力的 Agent 预设。
- [Graph](/zh/docs/the-one-app/definition-space/graph): 用节点和边搭建可运行、可观察的 Multi-Agent Graph。
- [Memory](/zh/docs/the-one-app/definition-space/memory): 每个 Definition Space 都带有一份由 Agent 维护、随它迁移的记忆。
- [Atlas](/zh/docs/the-one-app/definition-space/atlas): 一张用来漫游和编写整个 Definition Space 的可视化画布。
- [运行](/zh/docs/the-one-app/run): 选择工作区路径和 Graph，启动一次透明的多智能体运行。
- [设置](/zh/docs/the-one-app/settings): 管理账号、Definition Space、API Provider 和桌面运行环境。
- [仓储 Space 管理](/zh/docs/the-one-app/settings/space-management): 创建、切换、导入、导出和分享不同的 Definition Space。
- [API 配置](/zh/docs/the-one-app/settings/api-configuration): 配置官方模型服务、第三方 API Key、OAuth 和 OpenAI-compatible Provider。
- [单一 Agent](/zh/docs/multi-agent-architecture/single-agent): 理解一个 Agent 由模型、工具和上下文共同决定。
- [模型](/zh/docs/multi-agent-architecture/single-agent/model): 模型是 Agent 的大脑，不同 Agent 应该选择适合角色的模型。
- [工具](/zh/docs/multi-agent-architecture/single-agent/tools): 工具让语言模型真正成为 Agent，工具边界就是能力边界。
- [上下文](/zh/docs/multi-agent-architecture/single-agent/context): 用约束、信息和任务三类上下文建立清晰的 Agent 心智。
- [ReAct](/zh/docs/multi-agent-architecture/single-agent/react): The One 中 Agent 默认采用推理、行动、观察再推理的运行方式。
- [行动链（Action Sequence）](/zh/docs/multi-agent-architecture/single-agent/action-sequence): 用行动链把复杂目标拆成更可执行的阶段路径。
- [从单一 Agent 到 Multi-Agent](/zh/docs/multi-agent-architecture/evolution): 解释为什么复杂任务需要多个角色清晰、上下文干净的 Agent 协作。
- [Multi-Agent 的更多可能](/zh/docs/multi-agent-architecture/more-possibilities): Multi-Agent 相比单一 Agent 多出的独特能力边界。
- [通讯方式](/zh/docs/multi-agent-architecture/more-possibilities/communication): The One 中 Agent 之间的三种主要通讯方式。
- [上下文编辑](/zh/docs/multi-agent-architecture/more-possibilities/context-editing): 允许 Agent 在 Graph 中修改其他 Agent 的上下文。
- [Dynamic Graph](/zh/docs/multi-agent-architecture/more-possibilities/dynamic-graph): 让 Agent 在运行过程中动态创建、克隆、激活或休眠其他 Agent。
- [Meditation](/zh/docs/multi-agent-architecture/more-possibilities/meditation): 让 Agent 自己管理上下文窗口，压缩自己的历史。
- [Self-Evolution](/zh/docs/multi-agent-architecture/more-possibilities/self-evolution): 一个纯 Agent 的工坊，诊断自己的失败并在正确的层级上修复。