Agent

Overview

An Agent is an AI worker you configure with a role, knowledge, and tools. Agents act as the "brain" executing logic within your workflows — you recruit them to perform specific roles (e.g. Customer Support, Inventory Analyst).

Recruit an agent

Go to Agents in the side menu and click Create. You'll go through these steps:

1. Setup

• Agent Type — pick the kind of agent you're building (e.g. support, analyst-style types offered by your plan).
• Model Configuration — choose an AI provider and model.
• System Context Prompt — describe the agent's role and behavior in your own words.
• Vertical Features — turn on any relevant feature toggles for your use case.
• Manage Login — turn on if users must authenticate before talking to this agent.

2. Embed Settings

• Agent Name (required)
• Agent Description (required)
• Avatar — pick an icon.
• Primary Colour — pick the agent's accent color.

3. MCP Servers

Browse available MCP servers and enable the specific tools you want this agent to use from each one.

4. Additional Resources

Upload reference documents (PDF, DOC, DOCX, TXT, CSV, JSON) the agent can use to answer questions or make decisions.

5. Additional Prompts

Add extra prompt instructions that get appended to the system prompt — useful for tone, edge cases, or extra rules.

6. Custom Functions

Under "Configure from Scratch," click Add Custom Function to define a name, description, and code for a function the agent can call.

7. Share

Choose how the agent will be embedded:

• Widget — a floating chat widget.
• Page — a full page.

Copy the generated embed script and paste it into your website.

:::tip Fill in Setup and Embed Settings first — Agent Name/Description and the system prompt shape everything downstream, including how the agent behaves when used inside an Autotask. :::

Next steps

Use your agent inside an Autotask workflow, or attach it directly to an Automation.