Your AI Agents
The Your AI Agents page is the main section of the Admin Dashboard. This is where you create, manage, and publish all of your AI agents.
Navigate here via the Agents item in the navigation drawer.
The agents list
Each agent appears as a card showing:
- The agent's name and its system prompt description.
- Any tools currently attached (shown as labeled chips).
Actions on an agent
Each agent card has two action controls:
- A visible Share button — opens the share dialog directly.
- A View more dropdown — expands to reveal four icon actions:
| Action | What it does |
|---|---|
| Share | Opens the share dialog (same as the Share button). |
| Configure Tools | Opens the Tool Configuration page for this agent. |
| Edit | Opens the agent editor. |
| Delete | Permanently deletes the agent. |
Creating a new agent
Click the + button in the top-right of the page header (there is no text label — it's an icon button). A dropdown appears with two options:
- Create from Template — opens the template selection dialog. Pick a pre-configured starting point and click Continue to load it into the editor.
- Create from Scratch — opens a blank editor.
There is also a refresh icon button (circular arrow) beside the + button, which reloads the agent list.
The agent editor
Clicking Edit (or Create from Scratch / Template) opens the agent editor. The editor header shows:
- The page title (Create New Agent or the agent's name when editing).
- A subtitle: "Configure your new AI agent with custom instructions and tools."
- A Back to all agents breadcrumb link.
- Two toggle icon buttons (top-right) that switch between Basic and Advanced mode.
Basic mode
The default view. A form organized into three sections:
1. Basic Info — Name, model, and prompt
| Field | Notes |
|---|---|
| Agent Name | Required. What end users see. Typing a name and clicking out may generate a system prompt auto-suggestion. |
| Text Model | The LLM for chat. Has a Compare button and a search box. See LLM (Model). |
| Speech Model | Optional voice model. Default shows "No speech model selected". Has a Compare button. |
| Thinking Budget | Off / Low / Medium / High. Only appears when a thinking-capable model (e.g., Claude) is selected. |
| System Prompt | The agent's instructions. Placeholder text: "What should this agent do? Include tone and any must-do steps." See System Prompt. |
2. Tool Integration
A list of tools attached to the agent. Shows "No tools configured" when empty.
- + button — adds a tool. Opens a dropdown: Pre-defined Tool (catalog) or From Scratch (blank form).
- ? icon button — opens the More Tools dialog with external links to discover MCP servers (Claude documentation and mcpservers.org).
See Tools.
3. Knowledge Base Integration
Subtitle: "Connect your agent to custom knowledge sources. One knowledge base per agent."
This section is always visible in the editor. Fields:
- Knowledge Base Name
- Description
- Add Documents button — upload files (PDF, DOCX, TXT, MD, HTML) or add web URLs.
The knowledge base is created when you save the agent. See Knowledge Base.
Each agent supports one knowledge base only. Plan your agent architecture accordingly — if different audiences need different content, create separate agents.
Advanced mode
Switches the editor to a JSON view. The section is labeled Agent configuration and has two toolbar buttons: Copy JSON and Format JSON (icon-only).
The default JSON when creating a new agent is:
{
"name": "",
"textModelId": "global.amazon.nova-2-lite-v1:0"
}
See Advanced Mode for the full schema and usage guide.
Saving
The button at the bottom of the editor is:
- Create Agent — when creating a new agent.
- Update Agent — when editing an existing agent.
If validation fails (empty name, unsaved tool entry, invalid JSON), an inline error appears and the save is rejected.
After saving, existing conversations with the agent continue using the previous configuration. New conversations use the updated one.
Deleting an agent
Delete is permanent. The agent record is removed; end users can no longer open it. Conversations they had with the agent remain in their own history.
Next steps after creating an agent
Creating and saving the agent is the first step. Depending on what you added:
| What you added | What to do next |
|---|---|
| Remote (MCP) tools | Go to View more → Configure Tools (wrench icon) to authorize and fetch the tool list. See Tools. |
| A Knowledge Base | Watch the KB status chip in the editor. It starts at Setting Up and moves to Configured once ingestion is done. See Knowledge Base. |
| Neither | Test and share. Click Share on the agent's card. |
Read next
- LLM (Model) — picking the right model.
- System Prompt — writing the agent's instructions.
- Tools — attaching tool integrations.
- Knowledge Base — connecting documents.
- Sharing — sharing with end users.
- Create an agent — full step-by-step walkthrough.