Skip to content
Karyam

Create Your First Agent

Agents are the core building blocks of Karyam.

An Agent is more than a chatbot.

It can:

  • Understand requests
  • Retrieve enterprise knowledge
  • Use tools and integrations
  • Execute workflows
  • Collaborate with humans
  • Trigger business actions

In this guide, we’ll create an HR Assistant that can answer policy questions, access company knowledge, and initiate approval workflows.

Video Tutorial

Section titled “Video Tutorial”

Follow the complete walkthrough to create your first AI Agent in Karyam.

How Agents Work

Section titled “How Agents Work”
User Request
      ↓
System Instructions
      ↓
Knowledge Retrieval
      ↓
Skills & MCP Tools
      ↓
Reasoning
      ↓
Workflow Execution
      ↓
Response

Step 1 — Open the Agent Builder

Section titled “Step 1 — Open the Agent Builder”

Navigate to:

Workspace → Agents → Create Agent

You will see the following sections:

  • General Information
  • Model & Tuning
  • Skills & Tools
  • Chat Settings
  • System Instructions

Step 2 — Configure General Information

Section titled “Step 2 — Configure General Information”

Name

Section titled “Name”

Choose a descriptive business-oriented name.

Good examples:

  • HR Assistant
  • Customer Support Agent
  • Invoice Processor
  • Vendor Onboarding Agent

Avoid generic names like:

  • Bot
  • Assistant
  • Agent 1

Description

Section titled “Description”

Describe:

  • What the agent does
  • Which systems it can access
  • Which business process it supports

Example:

Handles employee leave requests, HR policy questions,
and approval workflows.

Step 3 — Select a Model

Section titled “Step 3 — Select a Model”

Choose the model that best matches your use case.

Use Case Recommended Model
General reasoning GPT-5
Enterprise workflows Claude
Cost-sensitive tasks Gemini Flash
Private deployments Ollama

Max Steps

Section titled “Max Steps”

Controls how many reasoning steps an agent can take.

Scenario Recommended
FAQ Agent 5
Workflow Agent 10
Multi-step Reasoning 15

Higher values improve reasoning capabilities but increase latency and cost.

Temperature

Section titled “Temperature”

Controls creativity.

Value Behavior
0.0 Deterministic
0.2 Highly focused
0.5 Balanced
0.8 Creative
1.0 Highly creative

Recommended values:

  • Support agents → 0.2
  • Workflow agents → 0.1
  • Creative writing → 0.8

Caching

Section titled “Caching”

Enable caching to:

  • Reduce costs
  • Improve response times
  • Speed up repeated queries

Recommended for:

  • Knowledge assistants
  • Support agents
  • Internal copilots

Step 4 — Attach Skills and Tools

Section titled “Step 4 — Attach Skills and Tools”

This is where agents become capable of taking actions.

Sub-Agents

Section titled “Sub-Agents”

Delegate work to specialized agents.

Example:

HR Assistant
├── Payroll Agent
├── Recruitment Agent
└── Leave Management Agent

Skills

Section titled “Skills”

Skills allow agents to interact with systems and perform actions.

Examples:

  • Send Email
  • Search CRM
  • Query Database
  • Create Ticket

MCP Tools

Section titled “MCP Tools”

Connect external systems using the Model Context Protocol.

Examples:

  • GitHub
  • Slack
  • PostgreSQL
  • Jira
  • Salesforce

AI Flows

Section titled “AI Flows”

Flows allow agents to trigger business workflows.

Examples:

  • Employee onboarding
  • Vendor approvals
  • Incident escalation

Step 5 — Configure Chat Settings

Section titled “Step 5 — Configure Chat Settings”

Voice Input

Section titled “Voice Input”

Enable voice conversations by attaching a voice input skill.

This allows users to interact with the agent using speech instead of text.

Ideal for:

  • Mobile users
  • Field teams
  • Hands-free environments

Step 6 — Write System Instructions

Section titled “Step 6 — Write System Instructions”

System Instructions define:

  • Persona
  • Responsibilities
  • Constraints
  • Communication style
  • Decision boundaries

This is the single most important configuration in an agent.

Example:

You are an HR Assistant for Acme Corporation.


Responsibilities:
- Answer HR policy questions.
- Assist with leave requests.
- Trigger onboarding workflows.


Rules:
- Never invent company policies.
- Always cite company knowledge.
- Escalate uncertain cases to HR staff.

Example Configuration

Section titled “Example Configuration”

HR Assistant

Section titled “HR Assistant”
Setting Value
Name HR Assistant
Model Claude Sonnet
Temperature 0.2
Max Steps 10
Skills Email, HRMS Search
Flows Leave Approval
Knowledge Base HR Policies

Best Practices

Section titled “Best Practices”

Keep Agents Focused

Section titled “Keep Agents Focused”

Prefer multiple specialized agents over one giant agent.

Use Knowledge Instead of Instructions

Section titled “Use Knowledge Instead of Instructions”

Store changing business information in knowledge bases rather than system prompts.

Use Human Approvals

Section titled “Use Human Approvals”

Require approvals for:

  • Financial actions
  • External communication
  • Sensitive operations

Enable Observability

Section titled “Enable Observability”

Monitor:

  • Agent Logs
  • Workflow Logs
  • RAG Logs
  • Token Usage

Production AI requires visibility.

Your First Agent Is Ready

Section titled “Your First Agent Is Ready”

Congratulations 🎉

You have created your first AI Agent in Karyam.

Next, connect enterprise knowledge and business systems to unlock the full power of AI systems.