Configuration and administration
Create and configure a data agent
Connect data and configure business knowledge to create a data analysis assistant for your team.
Guide · Module: agent · Users · Deployment: Not specified · Confirmed product versions: Not specified · Source: Existing documentation
A data agent is a data analysis assistant you configure in AskTable. After connecting business data and adding metric definitions, colleagues can ask questions in natural language and have it query data, perform calculations, and explain results.
For example, create an agent for store operations: bind sales data, define revenue and returns, and attach year-over-year and period-over-period analysis skills. The team can then ask questions using the same data and methods without repeating the background every time.
What it can do for you
- Find the tables and fields relevant to a question, then generate and execute queries.
- With the corresponding capabilities enabled, calculate with Python, visualize results, or ask for clarification when a question is unclear.
- Search public information as analysis context when an administrator has configured web search and the agent has it enabled.
Available actions depend on the capability switches and authorized data scope.
How data, skills, and entry points work together
| Supporting resource | What it addresses |
|---|---|
| Data sources | Which business data can be analyzed |
| Business documents and knowledge preferences | How metrics are calculated, what business rules apply, and how answers should be organized |
| Skills | Which method to use for a type of question; the agent reads a skill's body when needed |
| Channels | Ask questions in Feishu or Slack; only one platform can be enabled per agent at a time |
| Scheduled tasks | Run analysis and deliver results on a schedule |
| Web embedding | Provide a data Q&A entry point on your own pages |
These entry points share the agent's analysis configuration. A user's accessible data is also constrained by the entry point's user authorization or bound role.
Parts of an agent
Open an agent to see 7 tabs at the top. Configuration belongs under “Configuration”; external entry points and execution records have their own tabs.
| Tab | What it manages | Details |
|---|---|---|
| Configuration | Data source bindings, capability switches, skills, domain knowledge, and analysis preferences | Basic settings and knowledge preferences |
| Documents | Persistent business rules and definition documents | Business documents |
| Channels | Choose Feishu or Slack; configure credentials, reply behavior, and user permissions | Feishu · Slack · Feishu user permissions |
| Scheduled tasks | Analyze and deliver results automatically on a schedule | Scheduled analysis reports |
| Embed | Embed in your own webpages and control visitor scope | Web embedding |
| Audit logs | Execution trace and status of each question | Logs and user feedback |
| User feedback | Thumbs-up and thumbs-down ratings for tuning | Logs and user feedback |
The four capability switches under Configuration determine what the agent can do and are a central part of setup. See the dedicated page: Capability switches explained.
Before you start
- Permissions: your project role must be Project owner or Project administrator. Ordinary project members can ask questions but cannot modify this configuration.
- Prerequisite: the project has at least one data source. Otherwise, use main navigation on the left → “More” → “Data sources” → “Add data source” at the top right.
- Entry point: main navigation on the left → “More”, which opens Agents directly. You can also click “Agents” in the data workspace sidebar.
Create an agent
Open the agent list
Main navigation on the left → “More”. The page is titled “Agents”, with a “Create” button at the top right and a search box on the left for finding agents by name.
Choose a template or start from scratch
Click “Create” to open “Create data agent”. Choose a template or start with a blank agent.
- Template: each card shows a name and one-sentence description. Clicking it creates the agent with its name, description, and skills preconfigured.
- “Create from scratch” lets you customize the name and description and configure everything yourself. Enter a required name and optional description, then click “Create”.
Connect data sources
Creation opens the detail page directly. Select “Configuration”, then “Add” under “Data sources” to open “Select data sources”. Check the sources and close the dialog. See Basic settings and knowledge preferences.
Add capabilities, skills, documents, and preferences as needed
Review the capability switches, attach skills, and add business documents and domain knowledge and analysis preferences.
Check the list
Click “Agents”, the first breadcrumb on the detail page, to return to the list. Cards show the name, description, data source engine tags, and relative creation time, such as 3 days ago.
Built-in agent templates
The creation dialog offers 9 templates. Templates are read-only starting points; agents created from them can be freely modified.
| Template | Description |
|---|---|
| Store Operations Analyst | Focuses on retail store operations data to uncover opportunities and problems |
| E-commerce Data Monitor | Monitors core e-commerce data in real time to spot anomalies and opportunities promptly |
| Financial Data Analyst | Analyzes financial data in depth to understand business performance and identify financial risks |
| Market Insights Analyst | Tracks market changes, analyzes competition, and identifies opportunities |
| Supply Chain Monitor | Monitors every supply-chain stage in real time to identify risks and optimization opportunities promptly |
| User Growth Analyst | Tracks user growth metrics, evaluates growth strategies, and identifies opportunities |
| Executive Data Assistant | Provides one-click business data summaries to support fast executive decisions |
| Business Traffic-light Analyst | Sets red/yellow warning rules for business lines and automatically tracks metric health |
| Data Quality Guardian | Proactively monitors data quality to ensure accurate, trustworthy analysis |
Key constraints
| Item | Value or default | Where to change it |
|---|---|---|
| Data sources | One or more can be added | Details → “Configuration” → “Data sources” → “Add” |
| Python interpreter | On by default | Details → “Configuration” → “Capabilities” |
| Chart visualization | On by default | Details → “Configuration” → “Capabilities” |
| User clarification | On by default | Details → “Configuration” → “Capabilities” |
| Web search | Off by default; disabled until configured by an administrator | Details → “Configuration” → “Capabilities” |
| Domain knowledge and analysis preferences | Maximum 1024 characters each; save on blur | Details → “Configuration” → “Preferences” |
| Built-in agent templates | 9 read-only starting points; created agents are freely editable | “Create” dialog |
| Channel allowlist | Independent for each platform and off by default; when off, any user who can contact the bot can ask data questions | Details → “Channels” → “User permissions” (enable the platform first) |
Troubleshooting
| Symptom | How to identify it | What to do |
|---|---|---|
| Save is rejected | Effective project role lacks configuration-management permissions | Ask the project owner or administrator to change the configuration or authorization |
| Web search switch is disabled | Description says web search is not configured and an administrator must configure it in system settings | Ask a system administrator to configure it, then return to enable it |
| Question is rejected because no tables or fields are queryable | Agent's Data sources section is empty | Add sources under Details → “Configuration” → “Data sources” → “Add” |
| Question says table X in schema Y is missing | The table is outside the accessible scope of the agent's bound sources | Confirm the source is attached; if a policy blocks the table, change the rule under Policies |
| Conversation says its bound agent was deleted | The conversation's agent no longer exists | Start a new conversation or check the Agents list for the intended agent |
| Conversation says its bound role was deleted | The conversation's role no longer exists | Select another role or start a new conversation |
Next steps
- Understand the four switches: Capability switches explained
- Turn repeated analysis steps into a skill: Analysis skills
- Help the agent remember metric definitions: Business documents
- Use people, roles, and policies to control visible databases, tables, and rows: Permissions and data scopes