Product documentation

Create an agent

Create an Agent to answer questions with a configured model, instructions, and optional DataFact resources.

Prerequisites

You need corresponding access. At least one Agent creation path and a compatible model must be available in the workspace.

Create the Agent

In the sidebar, select the create action in the personal or team Agents section.
If DataFact shows multiple creation paths, select General agent or Data agent. If only one path is available, the form opens directly.
In Basic settings, enter the Agent name. Description, model, and tags are optional when the form already supplies valid defaults.
Review Resources and skills. This section opens by default because bound resources determine what the Agent can retrieve or analyze.
Expand Prompts and persona if you need to change the system prompt, identity, or response style.
Expand Other settings to review the answer language and output-display behavior that are available to you.
Save the Agent.

Expected result

DataFact confirms creation and opens the new Agent workspace. A Session is not created until you submit the first non-empty question. On your first visit to the creation form, a dismissible introduction explains that only the name is required and that the later sections can be skipped.

Important limitations

Available creation paths and resource choices vary by deployment and enabled capabilities.
A General agent requires at least one enabled skill. The creation form normally preselects enabled skills after they load; in a personal workspace, default skills can be visible but read-only.
A Data agent that uses a resource requires at least one resource of the selected kind. A no-resource Agent does not.
Team Agent creation is shown only to users with corresponding permission.
Creating an Agent does not configure recommended questions; that manager becomes available after creation.
Sections and controls that are unavailable in the current DataFact environment are disabled or omitted. Do not treat a missing control as an instruction to use an administrative page.

Errors and recovery

If saving is unavailable, check the name and every field marked as required. If a model, skill, or resource list is empty, that dependency may not be available in the current workspace.