Create an AI agent
Create a separate agent for a new task in the correct workspace.
Open copy menu
Create a separate AI agent when another use case needs its own task, knowledge base or conversation interface. If this is your first setup, follow the first-agent guide .
Before creating it, check the workspace, your editing permissions and the number of agents included in your subscription. If creation is blocked, check the reason with the workspace administrator before creating another workspace.
Create an agent and check its workspace
Section titled “Create an agent and check its workspace”Choose a workspace and create the agent
Open the workspace that should own the agent in Oma Aihio. Start creating a new agent and give it a name describing its task. If you use guided onboarding, add a source relevant to that task.
Check Overview
Open the created agent. Overview shows its model, data sources, instructions and operating status. If the knowledge base has no ready content, add a source and update it before evaluating answers.
Save and test one task
Describe the agent’s task and boundaries in instructions. Save changes and ask a preview question whose answer you can verify from the source. Creation has succeeded when the agent appears in the correct workspace and you can continue editing its settings.
Get to know the agent workspace
Section titled “Get to know the agent workspace”| View | What do you do here? | When do you open full settings? |
|---|---|---|
| Overview | Select a model, check knowledge status, edit instructions and manage operating status. | For advanced AI settings or instruction version history. |
| Appearance | Edit widget content, colors, the launcher and languages. | For separate lead collection or security settings, for example. |
| Actions | Review configured actions and where they can be tested. | To configure an API, credentials or a collection action. |
On large screens, the preview stays beside settings when you switch views. On small screens, switch between Settings and Preview. Test messages use real AI and consume message credits.
Name and template
Section titled “Name and template”Choose a name that helps your team recognise the task, such as “Product support”. Do not include customer personal information in the name.
Start with one clear task. An instruction template helps you write instructions; it does not automatically enable integrations or human support.
Customising the template
Section titled “Customising the template”- Add the knowledge sources needed for the task.
- Describe what the agent helps with and what it should do when information is missing.
- Adapt the conversation interface’s visible text and appearance to the use case.
- Check an answer to a known question and to one the sources cannot answer.
Creating an agent does not install it on your website. Continue to publishing after checking the content and test answers.