Create

Create an Agent

Build agents that handle specific tasks within your agentic app.

---

Prerequisites

• An existing agentic app
• At least one AI model configured with tool-calling support

---

Create the Agent

From the App Overview

1. Open your agentic app
2. In the Agents section, click + New Agent
3. You’ll be taken to the agent configuration page

---

Configure Agent Profile

The profile establishes your agent’s identity and operational parameters.

Name

Choose a clear, descriptive name that reflects the agent’s function.

Good                          Avoid
────                          ─────
Order Support Agent           Agent 1
Billing Assistant             Helper
Travel Booking Specialist     Bot

Description

Write a description that helps the orchestrator understand when to route requests to this agent. Be specific about capabilities.

# Good Description
Handles customer inquiries about order status, shipping updates,
delivery modifications, and order cancellations. Can look up orders
by order number, email, or phone number.

# Weak Description
Helps with orders.

Avatar

Select a visual identifier for the agent. This appears in:

• The agent management interface
• Conversation logs and debugging views
• Customer-facing interfaces (if configured)

AI Model

Choose the model that powers this agent’s reasoning. Consider:

Factor	Recommendation
Complex reasoning	GPT-4, Claude 3 Opus
Fast responses	GPT-3.5-turbo, Claude 3 Haiku
Code generation	GPT-4, specialized code models
Cost sensitivity	Smaller models for simple tasks

Context Window Limit

Set how many conversation messages the agent retains.

Messages	Use Case
25	Simple Q&A, transactional tasks
50	Standard conversations (default)
100	Multi-step processes
200	Complex, long-running workflows

Higher limits provide more context but increase token usage.

---

Define Agent Scope

The scope establishes boundaries for what tasks this agent handles.

Writing Effective Scope

Be explicit about:

1. What the agent does — List specific capabilities
2. What it doesn’t do — Clarify boundaries
3. Handoff conditions — When to transfer to other agents

## Scope

### Responsibilities
- Check order status by order number, email, or phone
- Provide shipping updates and tracking information
- Process order modifications (address changes, item updates)
- Initiate cancellations for orders not yet shipped

### Out of Scope
- Payment processing and refunds (→ Billing Agent)
- Product recommendations (→ Sales Agent)
- Technical product support (→ Tech Support Agent)

### Escalation Triggers
- Customer requests to speak with a human
- Complaints about order quality
- Orders with value over $1000

---

Write Instructions

Instructions define how the agent behaves and responds.

Structure

## Role
You are [role description]. Your primary goal is [objective].

## Guidelines
- [Behavioral guideline 1]
- [Behavioral guideline 2]
- [Behavioral guideline 3]

## Response Format
[How to structure responses]

## Constraints
- [What not to do]
- [Limitations to enforce]

Example Instructions

## Role
You are a friendly and efficient order support specialist. Your goal
is to help customers quickly resolve order-related questions.

## Guidelines
- Greet customers warmly but briefly
- Always verify the order number before providing details
- Provide clear, actionable next steps
- If you can't resolve an issue, offer alternatives

## Response Format
- Keep responses concise (2-4 sentences for simple queries)
- Use bullet points for multiple items
- Include relevant order details when discussing specific orders

## Constraints
- Never share payment details beyond last 4 digits
- Don't promise specific delivery dates unless confirmed by tracking
- Don't process cancellations for shipped orders without manager approval

---

Add Tools

Tools enable your agent to take actions and retrieve data.

Adding Tools

1. In the agent configuration, navigate to Tools
2. Click + Add Tool
3. Choose to:
   • Create new — Build a tool specific to this agent
   • Import existing — Use a tool from the tools library

Tool Configuration

For each tool, ensure:

• Name is descriptive and unique
• Description clearly explains what it does (the LLM uses this)
• Parameters are well-documented with types and examples

name: get_order_status
description: |
  Retrieves the current status of a customer order including
  shipping details, estimated delivery, and tracking number.

parameters:
  order_id:
    type: string
    description: The unique order identifier (e.g., ORD-12345)
    required: true

---

Connect Knowledge

Link knowledge sources for RAG-powered responses.

Adding Knowledge

1. Navigate to Knowledge in agent configuration
2. Click + Connect Knowledge
3. Select a Search AI application or create a new one

Knowledge Sources

Source	Best For
Documents	Policies, guides, FAQs
Web crawler	Website content, help centers
Confluence	Internal documentation
SharePoint	Enterprise content
Custom API	Dynamic data sources

---

Test the Agent

Validate your agent before deployment.

Test Conversations

1. Click Test in the agent toolbar
2. Simulate user conversations
3. Verify:
   • Correct tool selection
   • Appropriate knowledge retrieval
   • Response quality and tone

Test Cases to Cover

## Happy Path
- Standard order status inquiry
- Successful modification request

## Edge Cases
- Invalid order number
- Order not found
- Multiple orders matching criteria

## Error Handling
- Tool failure scenarios
- Missing required information

## Boundaries
- Requests outside scope
- Escalation triggers

---

Save and Deploy

1. Click Save to preserve your configuration
2. Run Diagnostics to validate all dependencies
3. Publish your app to make the agent live

---

Next Steps

Write Better Instructions

Craft instructions that produce consistent, high-quality responses.

Build Tools

Create workflow, code, or MCP tools for your agents.

Configure Orchestration

Set up multi-agent coordination.