# Core concepts How projects, data sources, agents, skills, and permissions relate. 使用指南 · 模块: getting-started · 使用者 · 部署形态：资料未说明 · 已确认产品版本：资料未说明 · 来源：编辑整理 · 按产品 4.2.76-1110（CLI 0.0.18）核对；私有部署的旧版本可能不包含近期变化。 In AskTable, members ask questions about **data sources** through an **agent**. The agent determines which data can be queried and how it is analyzed; the asker's **data role** determines which of that data is visible to them. ![Main objects in a project: data sources are used by agents; an agent includes selected data sources, skills, business documents, and capability settings; project members and channel people ask through an agent. The queryable data is the intersection of the agent data sources and the scope allowed by the asker's data role.](https://docs.asktable.com/img/diagrams/concepts-en.svg) An agent uses data sources, skills, and business documents; members and channel people ask through an agent. ## Organizations and projects - An **organization** is the unit for billing and administration and can have several projects. Organization members, enterprise sign-in, and plans are managed in the organization. See [Organizations and projects](https://docs.asktable.com/en/organization/overview). - A **project** is the boundary for collaboration and data isolation. Data sources, agents, skills, permissions, and conversations all belong to a project, and projects cannot see each other's content. ## Data - A **data source** is a database connected to the project, an uploaded file, or a workbook. After connection, the system reads the structure of tables and fields (metadata). See [Connect data sources](https://docs.asktable.com/en/data/connect). - **Table and field notes** describe the business meaning of tables and fields and are the main basis for the agent to understand data. See [Table and field notes](https://docs.asktable.com/en/data/semantics). ## Agents An **agent** is a configured question-answering assistant made up of: | Part | Purpose | | ------------------------------------------------------------------ | ---------------------------------------------------------------------------------------------- | | Data sources | The data the agent can query | | [Skills](https://docs.asktable.com/en/agent/skill) | Reusable analysis methods, such as attribution analysis or a weekly report format | | [Business documents](https://docs.asktable.com/en/agent/documents) | Business rules and terms, such as metric definitions | | [Capabilities](https://docs.asktable.com/en/agent/capabilities) | Whether it can draw charts, run Python, ask users for clarification, search the web, and so on | Members can create their own agents and share them with specific colleagues or the entire project. Sharing grants use only, not access to the internal configuration. See [Create and configure agents](https://docs.asktable.com/en/agent/overview). ## People and permissions A project has two independent kinds of permission: | | Project role | Data role | | ----------- | ---------------------------------------------------- | ------------------------------------------------------------------------------------------- | | Determines | Which features one can manage | Which data one can query | | Values | Project owner, project administrator, project member | Created by administrators, such as “East China sales” | | Composition | Fixed | Combines one or more **policies**, which define visible databases, tables, fields, and rows | The data a question can actually reach is the intersection of **the agent's data sources** and **what the asker's data role allows**. Creating an agent or receiving a shared one never expands your data scope. See [Permissions and data scope](https://docs.asktable.com/en/permissions). Askers in channels such as Feishu and Slack do not necessarily have AskTable accounts. They are recorded in the project as **people profiles**, and administrators authorize them to use specific agents. See [Feishu channel user permissions](https://docs.asktable.com/en/permissions/feishu). ## Working with results | Name | Description | | --------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Conversation | Ask and answer, with follow-up questions. See [Conversations](https://docs.asktable.com/en/conversation) | | Data table | A saved query that can be refreshed. See [Save, download, and share](https://docs.asktable.com/en/results) | | Dashboard | A board made of several data tables. See [Dashboards](https://docs.asktable.com/en/dashboard) | | Analysis canvas | Multi-step analysis that keeps intermediate results. See [Analysis canvas](https://docs.asktable.com/en/canvas) | | Automation | Scheduled reports and metric monitors. See [Scheduled reports](https://docs.asktable.com/en/integrations/scheduled-reports) and [Metric monitors](https://docs.asktable.com/en/automation/monitors) | Conversations and dashboards are visible only to their creators by default. Project owners and administrators can switch to “Manageable” to see all of them. ## Next steps - Ask your first question: [Ask your first question](https://docs.asktable.com/en/getting-started/quickstart) - Build an agent yourself: [Build an agent from scratch](https://docs.asktable.com/en/getting-started/build-agent)